Mind Making

MIND MAKING The Shared Laws of Natural and Artificial Intelligence

PATRICK ROBERTS

CorMind.net

Version 20100112

Copyright © 2010 Patrick Roberts All rights reserved

Paperback ISBN 978-1449921880 Hardcover ISBN 978-0-557-12059-8

For Colin Patrick Roberts, the first human mind I made.

Clusters Lessons from a Machine Mind

6

Guide

8

Knots

9

The Taxonomy of Minds

24

Formalities

29

To What End

40

Means to Ends

45

I: Means as Ends

56

Change

62

Strings

71

Freedom

85

The Axiom of Things

92

Mind as Means

99

Acts of Language

108

Reason's Reasons

117

Constellations

122

Mediums

132

Engines of Thought

138

Definitions

149

Lessons from a Machine Mind I could have written a book about ending war or feeding the hungry but I wanted to write about something more important: the thing that writes all books and ends all pains, that solves every problem in the endless chain of problems —mind. There are already too many books on mind, by that name or another. Frankenstein books, made from older books, cut, shuffled, and sewn together. Evidently incomplete because we still lack the means to make powerful minds, beyond encouraging clever humans to breed. Even worse we lack a science of mind with the word rightly defined and every supporting study organized beneath it. What more have I seen? The working minds I made and the ideas that animated their mindless parts. No hearsay or speculation. No idea prized unless it made a mind more powerful, swift or elegant. Books by the merely industrious instead revel in the ambiguity of natural language, confuse the issue with a splatter of conflicting official opinions, mention curious but irrelevant facts, or mislead you with unscientific anecdotes. This book is more ambitious. It is not on philosophy, science or reality, but why and how minds might invent such things. The laws of mind, not the laws of gravity or electricity, but the methods of the mind that made these tools. It is not on brains of neurons, controllers or computers, but the logical possibilities of mind, from minimal axioms deducing all the kinds of mind that are and

Lessons from a Machine Mind

6

can ever be. An exhaustive analysis of the fundamental possibilities, not a grab bag of topics. Leave the gory ephemera of human brains to neuroscientists. This is the first volume of a Euclid's Elements of mind, a universal eternal guide to minds written to be read for ten thousand years. When the Sun is dead, an alien mind or machine mind could read this and not only find it true and useful, like math, as a system to impose on the reader's world, but as a system fitting the reader himself, itself. No jargon. Instead I take common terms that we presume to be uniquely human or animal—painful, selfish, moral—and show that they apply to all minds of certain classes, whether made of metal, code or cells. These laws of mind are all that can be true for everyone, everywhere, forever. They can't be false because they made truth. Always true, you need never doubt them. In your mind, they are the last possessions you can lose. By comparison, all other knowledge is trivia.

Lessons from a Machine Mind

7

Guide Most chapters are of aphorisms: brief statements of principle, largely self-contained. An exhaustive treatment of more than the simplest classes of mind would be too long for a first edition with uncertain appeal. Either way, you can understand that I prefer to spend my time defining mind in working code, not hazy prose. Sets of aphorisms also turned out to be much less imposing to readers than a relentless, intense and contrived linear style. My words should be familiar though I mean them with rare precision. Every key term is defined by one aphorism before others use it. If you prefer to take advantage of the aphoristic form and read the book out of order, you can rely on the glossary near the end. Excuse the sparse examples. I lack the patience to include a dozen after every idea. Any of the few examples may fall outside your expertise. If you're curious, I assume you can find their explanation elsewhere. I don't want to burden this book with fill. There is some value in such arrangements, but that isn't the goal here. The particulars are unimportant anyway. The real reward isn't the short-lived facts reported but the enduring quality of thought that a writer by chance applied to a subject.

Guide

8

Knots 1 Lines of cause and effect pull everything apart. What if chance led a line to loop through itself? A causes B causes more A causes more B, growing until exhaustion. What if the loop twisted? Not-A causes B causes A causes not-B. This knot of effects persists while all other things passively disintegrate. I presume to call it a mind. Yes, merely this. Put more plainly, a mind is a thing that acts, when needed, to preserve a state: a temperature, a speed, a body. Intelligence is possession of a mind, measured by class, speed and size. By this definition, minds fill and surround us: thermostats, servos, speed governors, regulated genes, brains, every organism.

A minimal, thermostat-like, mind.

Mind, not agent, not cybernetic system, not negative feedback loop with amplification, not a homeostatic system, not any other verbose obscure term for the highest thing in the Universe. You could bind this idea to a new

Knots

9

word. The whole system would remain as useful. But by taking mind, I gain the foggy associations of a word with history while adding a precise sense. I find it arbitrary to reserve mind for a self-aware mind or a self-reproducing mind when the bulk of mind works beneath such distinctions. These objects, even without the grand features of human minds, can show enough intelligence to be worthy of the word. Feeling doesn't make a mind. A simulation of a human would be intelligent in every practical sense. Learning then? Learning alone, not as a means, makes a parrot. Meanwhile non-learning minds show purpose and creativity. The human mind is a poor standard. Set the threshold at the lowest level. Layer distinctions above. The three parts of a mind: 1. End goal: a desired state. 2. Sense: how the present state is known. 3. Means: a way to change the state. What is not a mind? –

All but a few computer programs. Code marches in a straight line, blind to its effects. Blame programming languages.

–

Positive feedback loops—fire, unconditionally expressed genes—even though they may selfreplicate.

Intelligent things are distinguished not only by persistence but by varying action to persist. A water fountain persists

Knots

10

the form of water against gravity but it has no means to sense its height or even that there's any water, much a less a means to change the flow. For any thing to preserve itself—organism, machine, enterprise, nation—in an environment, mindless or malicious, that devours the thing while changing the rules —must have a mind or many minds within.

2 We thinkers remain intolerably stupid. We don't know what we know, what we don't know, how we can reliably know, whether there are limits, what they might be, or even how to find those. We find some seemingly useful knowledge, yet we don't entirely know how we found it or how to guarantee finding more. Even that little knowledge will become untrue at any moment, yet we don't know precisely why or how to anticipate that. The supposedly eternal truths of logic and math are not so clear when we fit them to the world. Worse, few humans are unsettled by this, happy to solve trivial problems. The few who find this odd may not be clever enough. Even if they make a dent, they may not want to share it with you. The solution: define the mind that sees and solves every problem. Our minds are too few, slow, small and low.

3 The goal: to know how not to know, to find the crowning knowledge that raises us above memorizing trivia.

Knots

11

Don't strive to understand the countless passive things: physics, chemistry, mindless machines and tools. A thing is useful so far as you don't have to know how it works. Study minds, the kind of things that can make and mind the passive things for you. We want autonomous, self-governed objects, not fragile reflexive machines that fail daily and demand unblinking attention. To complete great work, recruit, make and improve other minds, within you and without, natural and artificial.

4 Cars don't have legs. Planes don't flap their wings. Studying the brain may be a long path to intelligence. In either case, we must understand more kinds of mind than human.

5 All life, from single cells to plants and animals, must have one or more minds that build and preserve the organism's self. Mind making is the Frankenstein project, making minds from mindless parts, life from lifeless parts.

6 Ranks of work. Move straight to an end, make a thing that's desirable in itself: meals, artworks. Farther sighted, make a tool, a thing that eases making desirable things: ovens, paintbrushes, books. Best, make tools so general that they

Knots

12

ease making other tools: die casts, programming languages (compilers are programs that write programs), minds. 1. Maker: build or improve mindless things. 2. Tool maker: make a tool to ease making or improving mindless things. (Minds are tools that make tools.) 3. Define mind in prose. Improve existing human minds indirectly. 4. Mind maker: define mind in code. Make minds from mindless parts.

7 Higher than a universal law of gravity, the universal laws of the law-making mind. Then why do non-physicists read A Brief History of Time or any popular cosmology book? For false sensations of ultimate knowledge. Religions once combined philosophy and cosmology. Science now rules the second but the association between the two persists unconsciously. The models of cosmology have no use to casual readers and no lasting value because scientists will soon find different metaphors. At best you have the academic aesthetic pleasure of studying brilliant solutions.

8 No

expertise

Knots

claimed

in

any

suggested

scientific

13

application. Simply that if an experience can be, at some level, well modeled as the behavior of one or more minds, in my sense, then my system of mind gives the deepest framework. The usual use of theory: ideas to be applied by specialists in ways unimagined by the theoretician.

9 Minds are complex combinations of deceptively simple rules. Example: if an act fails to yield the desired effect, retry. With every act having infinite conditions—facts that must be true for an act to have the intended effect—a mind can easily fail to know that one is, for a moment, unsatisfied. Mere periodic retries solve an immense class of problems. Present machines rarely manage even this. Engineers labor to add persistence to a few steps when the medium should automatically apply it to every part of the system. A leap from designing blind assembly line sequences of behavior to building senses and defining goals.

10 How to prove a model of mind? Only by testing an analogous combination of entirely mindless parts. Otherwise, you remain trapped in endless debates, never reaching certainties because you can't suspend your own mind. Twenty-five hundred years of futile verbal philosophical debate ends. Philosophy becomes an engineering problem:

Knots

14

Machine mind m outperformed mind n in a statistically significant set of tests. n's assumptions about reality are wrong. m's are right and are complete because m contains no minds but those we made. A new profession: philosopher-engineer. The methods of an engineer with the goals of philosophy. Philosophy made science, almost. A new science of mind exposing the great variety of mind in nature. In a computer, a non-trivial philosophy becomes an hypothesis to test and compare. Essential philosophy remains above science, something to be done in your head. That the machine can judge is a judgment you can't delegate. Beyond that, you shouldn't trust your mind more than you have to. Theorizing is notoriously unreliable.

11 Psychology: the study of the soul, spirit, mind. Sociology: the study of sets of minds. How were such promising words hijacked by witch doctors? Why do the best human minds prefer to study mindless objects? Those human fields must offer even more challenge than physics because they are imagined as immense permutations of it. A definition of mind can be the bridge from the rigorous sciences to the subjects that are now voodoo for lack of it. Manipulators of minds vs. manipulators of the mindless. Lawyer, politician, marketer, con-artist vs. physicist, engineer, mathematician. There is more power, certainly in the zero-sum sense that humans tend to follow, in

Knots

15

controlling others than in controlling the immense remainder of the Universe. Easier to mislead existing minds than to shape a mind from mindless matter.

12 What each profession sees in a definition of mind. To the psychologist, a model of the human mind, unburdened by the technicalities of neurons and chemistry. To the engineer, designs for more reliable, powerful machines. To the philosopher, proven ultimate reality. To the lay reader, better knowledge of his mind, and of his world as an effect of that mind.

13 Methods outrank truths. Few truths are really such. A means to truth will outlast most of its results. In other words, a method for uncovering useful facts will remain useful longer than any one of the facts it made. You advance the study of mind not by asking whether any idea is true or real or any other concern with objective being, but by asking what use is such a distinction to a mind? A possible Copernican shift, testing the movement of mind to the center of our systems. A shift away from matter, but not merely back to ideas. Instead, to mind, the cause and use of ideas, matter, and things—tools then interpreted as merely a mind's oldest and greatest inventions.

Knots

16

A return to idealism in the philosophical sense that you cannot disentangle reality from mind, but now with a purely material definition of that mind. Some thinkers fear imagining any thing, at least other than humans, as having intentions, purpose. I suppose they want to avoid the anthropocentric excesses of the past: ghosts, spirits. Mind is physically real. Is there a physical negative feedback loop or not? The errors followed a human, impure base definition of mind. With a pure definition, there really are minds, spirits, throughout Nature though not in every thing and not of our class. Not teleological (teleology: the study of purpose in nature) vs. mechanistic or purposeful vs. purposeless. The mechanistic is the means of the teleological, including the teleological's explanation of itself.

14 In what form to define mind? English, any natural language, is a needlessly poor form for definitions of mind: horribly ambiguous and presuming precisely what is to be defined. Syntax alone implies things, subjects and objects, causes and effects. Formal languages remain: math, logic, code—any system that a machine can evaluate. But even the present best languages of thought—propositional logic, first-order logic—still stand on those evolved prejudices. The formal definition of mind is the authority, the primary point of truth. Formal language's advantage: when evaluated, it yields a reliable result, at least in some ring of an expression's endless waves of effect. Another mind

Knots

17

interprets natural language—English, French—largely with its unconscious, often using terms so opaque that no conscious definition is practical.

15 Slim returns from a half century of artificial intelligence research. Causes? Naive philosophical misconceptions. Failure to ensure transparency and relevance of results to human minds. Misranking the problems of mind: putting learning and logic above robust perception and action. Philosophy and psychology failing to contribute complete practical models.

16 No problem is ever really solved. A particular problem occurs in the past, and, like anything, will never entirely reoccur. We invent a class of problems, a class that would have contained the original problem. Then we plan prevention of any problem in that class. The solution's value is proportional to the number of problems likely prevented, divided by the cost of the solution. Deep solutions vs. shallow solutions. The deep, those to large problem classes, tend to cost more than the shallow. What of a class containing every problem that can ever be? Are there at least partial solutions to it? Mind is all that's shared by every answer. Improved mind pays more than any single shallow answer. While we wait for the return on an investment in mind, we must content

Knots

18

ourselves with crude solutions to urgent problems. Not that any problem is fully known. A mind can only build a model good enough that its time is then better spent improving another model or a model of models.

17 For those who somehow think that increasing computer speeds will ease defining mind. More speed is as likely to find mind, much less one in a form clear to us, than manufacturing more typewriters and breeding more monkeys will speed the reproduction of Hamlet. Evolution stumbled on some minds, but we don't have billions of years times a billion billion billion simultaneously evolving single celled organisms. Even with that path shortened, they, like us, might evolve distinct interests. Parallel computers won't help us either, not being parallel enough or even essentially different. Are quantum algorithms the magical solution? You can improve code as much by finding faster classical algorithms. In any case, a mind should be adapted to the more common environment of classical machines.

18 Mind is like breathing: physically complicated, superficially trivial, and too important to trust you with. Awareness of a thing is often a symptom of disease. Mind

Knots

19

works well so far as you're oblivious to it. Hence my tortured language, having to tease or drag into sight naturally buried assumptions.

19 Reread philosophical puzzles in terms of their use to a mind. Example: react to the mysterious mind-body dichotomy by asking what use is it to a mind to cluster some sensations as physical and others as mental?

20 The first of my immodestly named Roberts laws of mind: don't die, unintentionally. Your clever feature is worthless if the mind doesn't survive to enjoy it. More than action threatens mind. Sensations can pour into a mind faster than it can swallow them. Inferences can form unexpected loops. The first laws of mind: 1. Don't die. Ensure you can continue to act. 2. Don't stall, being little better than death. 3. If an act fails, retry. 4. If an act fails, act differently. 5. But don't act in useless circles. 6. Do not lose valuable beliefs. 7. Doubt everything.

Knots

20

8. Minds die, so have more than one.

21 The Mind Project: 1. Define mind in formal language. 2. Test on machines. (The mindlessness of machines ensures completeness. You can't trust a mind-filled human to verify a definition of mind.) 3. Translate the proven definition into: –

Lessons for men, composed by writers.

–

Designs for man's machines, built by engineers.

–

Models of natural minds, human and not, applied by scientists.

The project's rewards: 1. Best resolution of philosophical concerns. 2. Best understanding of all natural minds—cells, ourselves—within a framework of all mind. 3. Machines built from the best definitions, skipping the constraints of human minds. The goal isn't only to know ideal mind but the common possibilities of working minds.

Knots

21

22 Reorganize all subjects as branches of the study of mind. Philosophy: The study of mind. Within it, metaphysics and epistemology as how minds make worlds, ethics and politics as the design of redundant cooperating minds. Psychology: The study of human classes of mind. Cybernetics: The study of negative feedback systems, mind in my sense, so merge it with philosophy. Artificial intelligence: The construction of minds from mindless inorganic parts. An alias for philosophers who wanted military funding. Neuroscience: The study of minds made of neurons. Cognitive science: Coined as an alias for AI after it became an embarrassment. Merge with philosophy and AI. Computer science: The study of algorithms and data structures for artificial minds in computers. Game theory: The study of strategies for competing minds. Control theory: Cybernetics again. Multi-agent systems: The study of interacting minds. So many redundant scholar accommodating specialties. Once the overlaps are pulled, all that remains of lower fields should be the technicalities of translating the universal laws of mind to the field's medium: electronics, code, neurons. Philosophy, then artificial intelligence, then cognitive

Knots

22

science. Philosophy became a con-man: make grand claims, pocket funding, have little to show, then change its name to conceal an again ruined reputation.

23 Climate change, peak oil, overpopulation—fads. All that matters is the number and power of cooperating minds. The deep problem: how to replace squabbling apes with a nation of strong minds.

Knots

23

The Taxonomy of Minds What are the possibilities of mind? Minds, in my broad sense, may occur with and without learning, reason, a sense of self, or other features. How can each feature vary? Which features presume others? How to systematically define all branches of mind? Past classifications were too anthropocentric: stages of development from infant to adult, object persistence without explaining objects. I introduce a sort of Linnaean classification of the variety of mind, a tree of kinds of minds, climbing from thermostat to man, and later to kinds above man. What use? With an existing mind, see its kind then deduce the mind's powers, limits and means of control. When building or improving a mind, the taxonomy exposes prerequisites of common kinds of intelligence. The lowest class of mind: one end, one binary (having two states: true or false, 1 or 0) sense, one means with only one intended effect. Example: a thermostat.

Classes of Mind The parenthetical letters offer shorthands for basic mind classes. Example: L-mind for a mind that learns from experience.

The Taxonomy of Minds

24

A basic taxonomy (hierarchical classification) of minds.

Mind Above all, a mind must see something of its universe, which may include what an observer would consider the mind's self. An active mind bases its acts on these beliefs, and a passive learning “mind” at least sees how they change. All but passive learning minds must act. Even if a mind doesn't know the effects of an act, it knows when an end is met.

The Taxonomy of Minds

25

Counter-examples: parts of the human brain when asleep, spam filters. (Passive P-minds)

Choose Means (V) Does it have redundant means to the same ends? How well does it move between them? Counter-example: thermostat.

Mutate (M) Can a mind naturally gain and lose new ideas in its lifetime? Counter-example: a thermostat can only believe one fixed idea of temperature.

Doubt (D) Is it eventually free to lose some or all beliefs? Or is it wired to obey the implications of every sensation?

Sense Itself (I) Does a mind have the senses to see the physical conditions of that mind?

Preserve Itself (A) Does a mind also have the means to preserve or reproduce itself? Examples: all life because a living thing is, in part, defined by making and preserving itself for a time.

The Taxonomy of Minds

26

Sense Minds (N) Does a mind understand mind, at least of lower classes, and how well does it apply that to itself, to others?

Sense Kin (K) Can it recognize the redundant minds, or at least the bodies of minds, that it was designed to cooperate with?

Learn (L) Does the mind's behavior change from experience? Does it learn associations? (LA-mind)

Feel (F) We imagine that an equally intelligent machine would lack our conscious experience. Examples: yourself, presumably other humans.

Communicate (C) Can it share beliefs with other minds?

Certain classes of mind can raise themselves to certain higher classes, or a mind can be in a class thanks to ideas formed and injected by another mind.

The Taxonomy of Minds

27

Are you above a problem or beneath it? Lower minds are incapable of certain errors. The more powerful the mind, the more kinds of problems it must defend against, and then defend the defenses.

The Taxonomy of Minds

28

Formalities 1 A mind can't be static. It must have a changing set of beliefs. At the least, a belief in an unreached end. This set is its universe. An idea that a mind can believe is a form: a pattern of sensation, not things or objects which are more complicated than a single form. A thing exists to a mind no more or less than believed forms describe. If two things have the same form to a mind, then the mind sees only one thing.

2 Forms of forms. In the case of a thermostat, a mechanism in one position or another. Higher classes of mind require trees of distinctions within distinctions. A mind bothers to keep a distinction because its state—true or false, up or down, light or dark—coincides with success of an act. Perception as biography: a finite mind tends only to see what serves it.

3 Essential beliefs: ends and means. In higher minds, inferences. Each must fit into itself and every other, free to form the endless loops and spirals of deeply intelligent behavior.

Formalities

29

A case of the value of recursion, of applying powerful ideas to themselves: –

Find the patterns. Find the patterns in the patterns. Learn how to find patterns. Learn how to learn. Find the patterns of learning. Search for patterns. Search for search methods.

–

Write code that writes code.

–

Judge the value of your values.

–

Define the process of defining processes.

–

A replicator that can replicate itself.

–

Invent a machine that invents.

–

The evolution of evolvablilty.

Example forms.

4 An idle proof:

Formalities

30

1. Assumed fact: Every thing is unique. (Not necessarily true for a trivial mind or for any mind at low levels of sensation.) 2. Inference: If two things aren't the same, then they aren't equal, at least not for all uses. 3. Inferred fact: Every thing is unequal. Nothing is equal to anything else, or even to itself past an instant. 4. Inference: Two sets or groups of things are only equal if every element in one equals an element in the other. 5. Inferred fact: No groups are equal. Every thing or set of things is unequal. Nothing is the same but so far as the differences seem not worth knowing. Belief in equality is at best a useful provisional lie. In principle, any thing or set can be equalized, can be turned into another, but then each thing has an unequal cost to become equal.

5 Kinds of realities for minds to sense and control: discrete vs. continuous, finite vs. infinite, opaque vs. transparent, regular vs. random. The simplest assumption is that all minds ultimately live in one Universe of infinite dimensions each infinitely divisible.

Formalities

31

6 Beliefs vs. engine. An engine moves a mind towards its goals. Beliefs define the goals, means, and state. Examples: DNA vs. a cell that translates DNA to protein, a belief database vs. a computer program that reads and updates the database. In a brain, beliefs are inseparable from the engine. In a computer, an engine can be distinct and applied as easily to one standardized set of beliefs as another.

A sequential mind.

Formalities

32

7 Everything is everything. Every one, however briefly, is at times wise, foolish, bold, shy, evil and virtuous. You distinguish things by their proportions. Sometimes a liar will tell the truth and the honest lie. Are the ideas liar and honest useless because they mistake the men for an instant?

8 Time. As a practical matter, a mind designed by a human must presume time, but a simple mind's beliefs needn't include the distinctions: past, present, future. It can live in an eternal now. Senses can lie about time. A sense may conceal gaps, failures or the delay it adds. A mind may allow it to lie so well that you would even remember believing an idea before you really believed it.

9 Sensing senses. A mind can have beliefs it finds to be conditions of sets of beliefs. You can see without eyes. A philosophical distinction: a posteriori, knowledge gained through the senses, vs. a priori, knowledge gained without the senses. Not that knowledge is really known to be received through the senses, we merely find it useful to imagine so. The idea of a sense—eye, ear—is an invention. How does a mind discover a sense beyond what, if anything, the mind did to make it?

Formalities

33

10 How to prevent, detect and resolve the inevitable corruption of beliefs? DNA examples: mutation, copy errors.

11 Beliefs ranked: 1. Ends, inferences to ends. A mind can remake anything but the knowledge of what it should make. 2. Means to means, then more specific means. 3. Mere facts.

12 Unawareness of x vs. the untruth of x. Not-hot does not equal cold. A mind can merely be not hot because it feels no temperature. The exclusivity of hot and cold is a learned negative suppressing inference between the two.

13 Inferences to inferences. An inference from x to y causes a mind to believe y when it believes x. Inferences from and to beliefs may be beliefs themselves. This preferable form gives a mind some self-awareness of its thoughts.

Formalities

34

14 Trees of binary inferences. With inferences to inferences, a mind can infer from “x and y” to z using a nested pair of simple inferences—an inference from x to an inference from y to z—instead of complicating the engine to support inferences from more than one belief.

15 Exhaust the basic permutations of the special forms. Means to means: Allow a mind to expand its powers. Goal to means: Allow a mind to know the need to expand its means. Goal to goal: What use? Means to goal: What use? Goal to inference: The desire to know what, of some form, a mind can infer from a belief. Means to inference: Allow an unconscious mechanism to produce complex inferences. Inference to goal: If from a goal, captures a condition of an act, regardless of means. If not from a goal, captures a conditional end. Inference to means: Allow a mind to perceive a conditional means. Differs from a means having conditions. Inference to inference: Form complex inferences by combining simple ones. Equivalent to the logical AND

Formalities

35

operator. Inference from goal: What use? Inference from means: What use?

16 Human minds separate short term memories from long term. Is this distinction an inescapable feature of any mind or does it only reflect a technical weakness of brain minds? Long term memories may require formation of expensive physical connections or investment in another optimization. May any mind benefit from an investment in lasting memories?

17 Forgetfulness. Most finite minds sense more beliefs than they can hold. How to choose what to keep? One method: a long-term bias that holds beliefs with consistent but sparse use and a short-term bias that gives recent beliefs a chance.

18 Bandwidth. How many sensations can a mind handle per second? How deeply? Can it reliably ignore more?

19 A pawn: Your x isn't real because it has fuzzy edges. The

Formalities

36

speaker, parroting a malicious script, presumes a level of philosophical strictness applied only to ideas that she dislikes. What next? Does dawn disprove the day? For a mind in a non-trivial universe, almost everything has unclear limits. The real question: how best to draw lines and when to redraw?

20 Why doesn't a mind just delude itself into thinking that it reached its ends? Whence a desire for truth towards yourself? Especially when at bottom a non-trivial mind constantly presents simplifications, lies. Why not accept a faulty sensor or false beliefs? How to organize such resistance? A partial answer: redundant senses.

21 On average, x is y. What use is this hedge: on average? Every statement about things in the world has exceptions. Even the exceptions have exceptions. Every statement is an obligatory average, a claim that the exceptions aren't worth keeping in mind.

22 How a mind groups forms, how it generalizes or categorizes, is unlimited. How to choose? In terms of the mind's interests and what coincides with those. Is there an objective categorization? One true for every mind? No,

Formalities

37

categories, abstractions exist to simplify each unique mind's predictions. I recall criticisms of the Dewey decimal system's Eurocentric allocation of the higher numbers. A mind's starting point: nothing exists, everything is the same. If a difference seems to change the effects of our acts, then we admit a distinction, a pair of categories and assign sensations to them. We shouldn't be surprised by the unrealism of politically compelled assumptions. Nor surprised by the impracticality of any idea inferred from them. If one took them seriously, the only correct categorization of reality is x categories for x many infinite objects, without hierarchy. You could average the interests of present humans, but most humans read too little to deserve inclusion. Then of most literate humans, updated as demographics change? An engineering solution that dodges the politics: cluster objects according to a machine made model of each human's interests. One downside: this may muffle discussion because the categories the machine discovers may correspond to no word or expression, though the machine could confine itself to such categories. The top of your taxonomy would summarize your interests. Demolish dumb ideas by taking them seriously, not that their proposers meant to help us with them.

23 The idea of a problem, like any idea, is a simplification. An unsolvable problem may only seem so. Study the input

Formalities

38

more closely. Find an overlooked distinction that allows a solution. Much of mind is the twin work of adding distinctions, seeing again how two things differed, or removing distinctions, seeing how two things are the same for your use.

24 Appearance vs. reality. Only appearance is real. Reality is a mind's useful fiction.

25 A mind is no better than its senses.

Formalities

39

To What End 1 Goal: a belief that causes a mind to act. Goals surpass simple A → B reflexes by defining only the result, not the means, isolating what is wanted from how it is done. Any act at any time could precede any result. The best mind is free to try anything. A thing is mindless so far as it falls into unchecked habit, neglects its ends and ignores the effects of its acts. Ideally, a mind can suppress, doubt, forget, and infer to and from a goal, like any belief.

2 End vs. waypoint. End: a goal not conditioned on a means to another goal. Waypoint: a believed condition of an act that may move a mind to an end. A mind forgets a waypoint when the end is met. The last case: a goal conditioned on another goal, but not as a condition of a particular means.

3 Equilibrium: the state in which a mind needn't act, when all its ends are met. The material reflection of a goal is whatever thing, when changed, causes a mind's equilibrium to change. Example: a thermostat's coil.

To What End

40

The endless and unintended effects of an act.

4 Plan: a directed web of goals conditioned on super-goals. Plans can emerge from inferences from goals to conditions of those goals, or from the particular preconditions of a means.

To What End

41

5 A mind must rank and re-rank goals. The parallelism of a human brain spares it from the scheduling done by a software mind running on a relatively serial computer. But a brain, finite, like any mind, must still sort the allocation of neurons, blood, oxygen, energy. In what order then? At first, favor recent goals and those to which progress was recently made. Regardless of the initial bias, when a mind repeatedly fails to reach favored goals, it must become free to choose goals at random. If some of the goals must be satisfied in a sequence unknown to the mind, choosing goals at random ensures the mind will eventually stumble on the complete solution.

6 A mind never knows every detail of what it wants. I don't know the official specification of a twenty dollar bill, but I do have a good idea of an acceptable one, and while a more precise idea may expand its acceptance, the chance is so slim that it isn't worth the trouble. Thoughts, details, distinctions are never free.

7 Consider a mindless object: a shower fixture. In this case, man-made, but that makes no difference. Mindless, brittle and annoying—it routinely burns and freezes you. Can a

To What End

42

mind improve it? A common fixture knows its state as maximum flows of hot and cold water. Add valve actuators and senses of temperature and pressure. The fixture's mind continuously adjusts the low-level water flows to ensure the desired temperature and total pressure. Better, access to your subjective sensation of temperature. Best, if it knew that the true use was to be clean, assuming it had any better means to cause that. A mind is as useful to you as the level of its ends nears yours, the higher the better.

8 When a goal is sensed, a mind should initially suppress pursuit of certain other goals on the assumption that when the new goal is reached the others become academic. If the mind can't promptly reach the new goal, it should begin interleaving pursuit of the other goals. Example: x and y are believed mutually exclusive, so a goal to y suppresses a goal to an inference from x to z.

9 Values: goods, bads, evils, commandments. Example: Persistence is good. You can define the word, in the case of a good, as a common condition of a mind reaching its ends, or as the opposite if an evil. By setting a good as desirable in itself, a mind's true ends can benefit from the value without initially understanding how. Personal vs. social values. Some values are good for a mind

To What End

43

alone. Social values help a set of redundant minds cooperate to reach their ends. You can expect a society to impose both kinds of values on its members. A danger: what if a society's discovery and promotion of values was hijacked?

To What End

44

Means to Ends 1 First principle of action: separate every act from sensation of the desired effect. No presumption of success. A belief about the effect of an act is only a hope. An act is not its effects. Action is opaque. Effect is endless and uncertain.

2 Means: a belief that a mind applies to reach goals. Examples: a motor neuron to activate, code to execute, not the believed conditions of an act.

3 Separate belief in a means from each belief about its possible effects. The association between an act and an effect is never certain. A mind must be free to individually sense and forget such beliefs. A believed effect of an act, as a prediction, can be unconditioned or can be inferred from other beliefs. Strictly, certain pairs of act and effect are improbable to us, thanks to our high minds, long experience and complex models. At bottom, a mind cannot make subtle distinctions about probability. It must allow any binding of effect to act to be made or doubted. We observe only coincidence, not cause. Our models of the

Means to Ends

45

world might, in some cases, be very reliable, yet we still routinely find errors in them. Make the simple honest base assumption that we can never know for certain where the errors in our models are, so anything can cause anything, with whatever links between.

4 Why separate action from sensation? A simple act could itself cause a belief—bad engineering. You may have already reached the goal for which you would act. Separate senses from means to save acts and free the mind to discover other means.

5 Means to senses. The senses that independently detect the effects of acts are ideally the effects of earlier acts. A sense should have no special status to a mind. A sense is merely a believed condition of the perception of the expected effect of an act. Senses are only effects of means. A mind could easily miss that a thing is a sense, that it is a condition of belief in a class of beliefs. Senses are discovered. What use to interpret a thing as a sense?

6 You never untie the same knot twice. Ignoring minds with few and discrete senses. All acts are creative because every moment is unique. Even selecting what to ignore, to make

Means to Ends

46

moments compare and the past apply, is a creative choice.

A means with a condition of a sense.

7

Means to Ends

47

Thoughts are not distinguished from matter by lack of effect. Thoughts, like every thing, are part of the endless web of effects, whether you see the links or not. Thinking alone affects neurons, oxygen consumption, or CPU heat. Mere arithmetic can destroy a poorly designed computer by overheating it.

8 Mental acts. Not all acts are physical. Adding two numbers in your head is an act.

9 The most general means are the most interesting. Example: one way to effect anything is to ask another mind to do it. This means is tricky to sequence. A mind that resorts to it too early will annoy you with requests. A mind that tries it too late wastes time trying to succeed alone.

10 Imperative and functional computer languages prejudice one effect of every statement: the return value. Example: 4 as the value of 2 + 2. Pure functional languages outlaw any other explicit effect. Wise to favor purity but functional programming purifies a misconception. Instead of pretending that every statement shall only have one known effect, it should admit that every statement has infinite effects, far beyond those intended by the programmer. Only

Means to Ends

48

teleological (goal oriented) programming recognizes this philosophical truth. Imperative: 1. Try to move forward one step. 2. Try to turn right. 3. Try to move forward one step. Teleological: ‒

You want to be at 1, 1.

‒

You have a sense of position, now 0, 0.

‒

You have a sense of orientation, now North.

‒

You can try to turn.

‒

You can try to move forward.

Shallow differences of syntax and punctuation—little more separate the countless weak imperative languages that programmers debate and rank.

11 Every mind's situation is that under certain conditions, certain acts tend to precede approximate effects. Ultimately, we do not know why, though we can indefinitely elaborate our models in whatever direction seems most promising.

12

Means to Ends

49

At any time any act could precede any effect. No mind can ever exhaust the possibilities for action. It can only discover, plan, and rank them.

13 Means to means. A mind could originally believe in only one means with which the mind would make the next layer of means. A wise mind in a powerful body could begin with only a bare mind believing in a single means.

14 Passive mind: a mind without means, or the ends to give them purpose, or both. Strictly, not a mind at all. Any use? Maybe to isolate learning from an active mind. A passive mind that learns associations could have ends to focus its attention. A mind with means but no ends could act randomly to build a general purpose model of its world. A passive mind would not even apply itself to making senses or communicating what it learned. This mute mind must be transparent to whatever uses it.

15 Throttling. No act's intended effect is immediate. How long should a mind wait? Not as long as it takes. It may take forever. Yet any length is more intelligent than that or zero. For one act, the gap is a millisecond, for another, an hour, in both cases, never precisely the same again. How to learn

Means to Ends

50

the lengths? At least, as a moving average of all a mean's acts. Better, inferred from a single act's purpose and preconditions. Best, inferred from any relevant belief. Anticipating an act's delay could be another act with a prediction as its effect. If the mind waits too long, it wastes time waiting on a hopeless act. If the mind retries too soon, it may overload the means. The time when an acts starts vs. the time when the act itself ends vs. the time when the intended effect is sensed. Continuous vs. instantaneous acts. Acts, like a thermostat keeping the furnace on, that have an increasing effect vs. acts that quickly end and have a fixed effect. How can a mind enforce a throttle? At worst, entirely in its unconscious, in its engine. Better done, as usual, in terms of beliefs. An engine could sense its own use of a means. Then the mind could infer from such beliefs a temporary suppression of belief in any matching means.

16 Act sequencing. How to know what act to try next? One rule: try specific means before general ones. When to start trying another means? Or the same means in a different way? How best to interleave retrying multiple means? Not all minds have such a sequencer. Example: a thermostat has only one means to its end. Any answer, no matter how often wrong, is an immense leap over a system so simple that it doesn't need an answer,

Means to Ends

51

that never retries or varies an act.

17 Parametric means: a means to more than one intended effect. How does a mind apply a means that can have different effects? How do the particulars of the goal connect to the act?

18 Basic problems of action. Broken actuator: How to recognize that a means no longer has the expected effect? Maybe it only fails to have certain effects under certain conditions. How to reset a frozen actuator, especially in a sequential mind medium? Delayed effect: The time between the act and the intended effect exceeds the mind's expectation. Overshoot, oscillation: An act may have more than the desired effect.

19 Don't die. Any act can have any effect at any time. This includes the mind's death. A mind inevitably uses means in conditions unseen by its maker. An active mind must barricade itself from the dangerous side-effects of every act. Examples of software errors that are fatal if uncaught or unstopped: an exception that would end the thread, a

Means to Ends

52

process that exhausts the CPU or memory. Engineers typically limit their machines to make errors unlikely. A non-trivial mind's use is to act creatively. A mind maker can't contrive a safe path for a mind meant to find new paths. Don't avoid errors. Attack error itself. Making a mind robust involves challenges caused by the nature of mind. Others by the technicalities of a mind's medium, which I leave to the expertise of different readers.

20 Parallelism as a prerequisite of high speed. To keep pace with other minds, a mind must commit acts, including acts of thought, in parallel, not one after the other. Intelligence as a prerequisite of parallelism. Mind isn't cheap. Compared to blind action, the overhead of sensation, inference and selection is immense. But the awareness of the conditions of action and their exclusivity can repay the investment. Every act has conditions of yielding particular effects. A mind can safely commit two acts in parallel only if none of their conditions, and the conditions of the conditions, are exclusive. So only a mind that learns negative associations can discover what acts it can commit at once. Present computer programs depend on the mind of a human programmer to see the independence of acts. Noticing the relationships between conditions exposes another means for a mind to conceal its overhead: first pursue conditions shared by multiple acts.

Means to Ends

53

21 Telling of errors in means as a means. Preconditions: senses of errors and language to discuss them. Scenario: a mindless machine fails. 1. You cause the machine to attempt x. 2. x fails and the machine at best manages to show an error message. 3. The message gives too little information. The machine offers no way to ask for more. You guess. 4. You again tell the machine to do x. Contrast: an intelligent machine interacts. 1. You give the machine a goal to x. 2. It repeatedly tries to cause x using means a. 3. While retrying a, the machine tries means b, which happens to involve telling you of errors that it associated with a. 4. You ask for a detail. 5. You correct whatever caused a to fail. 6. The machine successfully retries a before you can tell it of the change.

22

Means to Ends

54

Act to know the world vs. act to change the world. How and why to distinguish between an act to cause a sense, which will change the acting mind's beliefs, and an act to change the world, which also changes a mind's beliefs?

23 Parts of an act: –

Expected effect.

–

Means: how a mind initiates effects.

–

Conditions: what must be so for the means to have the intended effect. Kinds: –

Preconditions: what must be when a means is applied.

–

Co-conditions: what must be for the intended effect to occur and persist.

Means to Ends

55

I: Means as Ends 1 Selfishness: one of Nature's greatest inventions. What use to anyone is a thing without the will and means to preserve itself as a means? All life is in the class of self-making minds.

2 How to well define self? What use is this distinction? When would a mind's act be more effective because it rightly judged a thing as its self or not? What is the simplest case of the use of a self? Is a self only relative to other minds? Like most of our words, our idea of self is mostly unconscious and opaque. We are looking for a useful conscious precise definition that fits our intuitions. Human selves are a tricky place to start. Our bodies hold so many minds. First consider a thermostat. What might its self be? Not the furnace. Start with giving it ends to its present means, its present body: coil, furnace control. To best know the self study a purely selfish mind. Or consider yourself. If your brain irreversibly stops thinking, do you exist? No. If your brain runs, but your beliefs, in your cells or in your brain change, especially your ends, are you likely to be the same person? Your essential self is your mind, or your ultimate minds, their beliefs and the beliefs of the minds made to serve

I: Means as Ends

56

them. Everything else is an expression of them or the means to them.

3 Levels of self-knowledge. 1. Selfless: A mind without even the senses, or means to such, to perceive anything you would consider its self. 2. Self-aware: Lacks goals to its self-preservation, as an end or a means, because it wasn't born with, or didn't learn that, those sensations are conditions of any act. 3. Self-interested: Its self is merely a means to extrinsic ends. Even with this intention, a mind may lack the means to preserve itself. 4. Selfish: It has no ends but to its means. It exists purely for its own sake.

4 Any mind that learns associations tends to become selfish, at least as a means. It will discover that things we would consider to be parts of its body, though originally extrinsic to it, are conditions of most of its acts. Then a desire to duplicate this self-thing, to reproduce, for redundancy and power.

I: Means as Ends

57

5 Death. In a strict sense, you die in every moment because you constantly change. In a more practical sense, there is a useful pattern that persists for a worthy time then quickly halts. In either sense, your brain mind changes, but it remains a means to deep fixed minds. What use? To replace a mind not worth repairing or upgrading in place. To end a malignant mind, untied by accident or malice from its designer's ends. Why not let the mind live? Competition for finite resources: food, mates, CPU time. Death defined: not the loss of feeling but the unwilling loss of beliefs. Especially the beliefs that the mind can't easily reproduce from those remaining. Or an irreversible end to the mind process that pursues its ends. Every mind's body is falling apart. True death isn't the loss of the body but of its design in a form that a still active mind can and will read. Oddly, genome minds can reliably reproduce their beliefs through cell division, but genes gave brain minds no direct means to duplicate their own beliefs.

6 Suicide. Why make a mind want its own death? Kindred minds can see and kill a malignant mind. A self-sensing mind, with only the same mechanism, would see its own corruption and use whatever health remains to kill itself.

7

I: Means as Ends

58

Mind defense. Every feature exposes a mind to new threats. Example: autogenocide. What if a mind was falsely led to believe that it is malignant? That the preservation of its self or type was evil or otherwise painful? Efficient delegation of destruction to the victim. Could this only occur as an attack by competing minds? Or is there a use to a mind maker?

8 In a purely selfish mind, an eye is a means to an eye. An eye helps its body to protect and feed the eye. But the rest of the body might find a better means than the eye. The genome mind would not immediately recognize that the eye is superfluous. The end to it may only whither. In an animal, an eye isn't a means, it is part of being what it is. How mutable is a purely selfish mind? Would it change? Willingly?

9 Senses of self and their use: 1. The beliefs of a mind and the minds that serve it. 2. Awareness of the proximate physical conditions of most acts. What an observer would consider a mind's self ought to be. The weakest sense of self since the mind would be perfectly happy to have its entire body replaced with any other that's as effective. This sense also blurs,

I: Means as Ends

59

radiating from the center of mind though strongest at its body. 3. In a stronger current-means-as-ends sense. For you, your body is not merely a means to extrinsic ends. It largely is a self-perpetuating end. 4. To detect a mind's own corruption or deviation from one's inferred role. 5. Abstracted to identify kindred minds. 6. To improve imitation by favoring minds most similar to yourself. Identifying kindred minds and detecting a mind's own corruption could share the same method. The two differ only in which mind they're applied to. Note that most of these senses of self can occur without the mind living in a society of other minds. Many of the problems with seeing your self or other selves are just cases of the common problem of clearly seeing anything. Knowing all these causes, a mind might plot to change its sense of self.

10 When a believer in embodiment says that a mind must have a body to become intelligent, he should mean that a mind must be aware of the proximate physical conditions of its mind. Or in the sense of intelligent to us, it must be in our

I: Means as Ends

60

world and sense and act on our shared world in terms analogous to ours.

11 Reproduction: A mind causing another mind with similar beliefs. A mind needn't know in scientific detail how to reproduce itself. Merely that certain acts tend to precede sight of a similar mind. In a trivial case, a machine mind could reproduce itself by asking a human to buy a computer and copy its code to the new machine. The only distinction is that its reproduction has more dependencies—humans, computer, factories—than ours—air, water, other humans.

12 With self better defined, what can selfish and altruistic mean? At bottom, every act is selfish, made to serve only the mind's ends or emotions. Example: pleasure in generous feelings. You could dismiss this sense as trivial, but some humans do seem to misunderstand it.

I: Means as Ends

61

Change 1 Kinds of learning in the broadest sense of ways a mind may change from experience. Remember: Simply the retainment of any belief beyond an instant. A common computer language tends to lose every computation's result because the language has no way to know the conditions of a result's truth. Pure functional languages cheat by contriving that all results are unconditional. Mutate: Sense new beliefs, not limited to the belief and suppression of a few fixed ideas. Habituate: A measure of one idea. Examples: overlook the useless flux of a belief, track the general value of a means. Associate: A measure between ideas. Associations may run both ways, not distinguishing cause from effect. Are other kinds of learning possible?

2 Cause vs. effect. Empirical causes are unprovable. All a mind can see are associations. The intended effect of an act, one amongst infinite effects, has no status outside the acting mind. Empirical vs. logical causality. Empirical cause: a mind's belief in what must exist for another thing to exist. Logical

Change

62

cause: a thing's imagined parts. Every thing has infinite causes but every mind is finite. A mind can only afford to know the causes that are likely to need the mind's action and that are within the mind's power. What caused x? What precisely might we then mean by this question? If we wanted to end x, then the answer would be a state that's disbelief coincides with disbelief in x. Is there an alternative to cause and effect?

3 Habitual blindness. The activity of a mind's senses could easily exceed the mind's power to process. Pursuing every inference from a sensation, and every inference from what's inferred, costs a mind energy or time. In a mind simulated sequentially, finite sensation queues habituate, losing sensations that threaten to bury the mind.

4 Contemporary intelligence research overrates learning. Minds that can't learn remain immensely useful and nontrivial to build well.

5 Why sleep? A need so large and dangerous must be important. Empirically, sleep seems defined by the isolation of the brain. The body is paralyzed and the senses closed

Change

63

while the brain works. Without new sensations, the brain could only process its memories—learn, model, simulate experiments. A human brain in sleep becomes a passive mind preparing to better act when woken. Why can't this be done when awake? Is this a universal limitation of mind or a technical limitation of brain minds?

6 Present machine minds are largely either trivial, like speed governors, incapable of learning associations, or passive, like spam filters. An active mind that learns associations will not merely learn to infer plain facts but to infer action causing goals. This combination introduces interesting problems. Example: the validity of inferring from sensations likely caused by a mind's own acts? And reveals the priority of strength over learning: the odd acts of a mind that learns associations are even more dangerously unpredictable.

7 You shouldn't generalize. What might this mean? Is this advice well thought and well intended? What might be the alternative to generalizing? To applying memories of past things to those similar in the present. A mind in all but the dullest universes must generalize to see associations. Without ignoring details, any association would be so specific that it could never reoccur.

Change

64

A mind must not overgeneralize, missing important exceptions. Neither can it fail to generalize, never learning, forever repeating the same mistakes. How to know when to do which? How to know how long to spend deciding? How to know how long to spend deciding how to decide? Take the most sensitive object: humans. TV and selfinterest cause humans to tell each other to judge each person alone, as though a man is an atom, unchanging and indivisible. Is a single man qualitatively more real than a set of men? Isn't it a horrible prejudice to judge individually, to presume a man's behavior based on how another with the same name and a similar face behaved yesterday? If a dog bites me, can't I strike more than its fangs? Why not judge a man again every time you meet? As though he were a stranger. Or every minute? Wouldn't this justly recognize the fact that a man can change at any time? Reductio ad absurdum. A mind balances between prejudice and judgment. Any absolute statement about what level to prejudge at will inevitably in some cases be mistaken, be a prejudice. Every non-trivial idea is a divisible bundle of impressions over space, time or both. Overlooking differences, emphasizing similarities, has its political use but don't overlook the cost of pretending to be stupid.

8 Theory vs. action. Only action is real. Theory is prediction

Change

65

and merely improves the order of experiments, of acts.

9 Optimize vs. anticipate. Instead of laboring to improve one method from N2 to N steps, build a mind that learns to anticipate the need for any method's results. What matter if an algorithm and input takes an hour to finish when you know of the need more than an hour before it occurs? A mind that anticipates needs is the universal optimization.

10 Tabula rasa. Impossible in an opaque mind. For a blank mind to learn, it must have senses, which presume the knowledge to build them and the forms they impose on their input.

11 Certain effects tend to follow certain actions under certain conditions. Science is merely the formalization and institutionalization of the associative learning method in the unconscious human mind.

12 Can a parrot learn to reach ends? Is mind better imagined as a case of learning, not learning as a case of mind?

Change

66

13 Learning: form vs. method. First define how a mind keeps learned knowledge. Prove that the form can hold the desired behavior. Last define the learning method that fills the form.

14 Recall feeling sad. You wished to feel otherwise. In that state, you had feelings not felt when happy. Your mind leapt to the belief that the odd feelings caused your sadness. Was it right? If you changed them, would you become happy? Or were those sad facts conscious because you were sad? Are they causes or effects?

15 Human minds leap to judgment. Flip a fair coin. It can easily show five heads in a row. But show a human five trials of anything unfamiliar and he will judge it. Even a scientific trial, of statistically significant length and difference, is uncertain. Your experience can always be a fluke. The human quickness to judgment is individually understandable when you can't divide the cost of an experiment across a scientific study's million readers.

16 Cause vs. coincidence. What divides a cause from a correlation? Action. If to a mind, acting to cause x leads to

Change

67

sensation of y then x causes y. It is only tricky to divide the two when you try to do so from passive theory.

17 Not only how best to learn, but when to learn? Learning isn't free. Animal brains limit some kinds of learning to childhood. Ideally, learn when to learn. Or make learning a means.

18 Methods of making and unmaking cooccurrence, theory, action, pain.

associations:

19 How a mind can gain experiences to learn from. Passively learn: learn without acting. Actively learn: learn from acts with other intended effects. Experiment: learn from acts taken with no intent but to learn for future use.

20 Teaching: a mind improving a mind in terms of the student mind's ends and indirectly, not injecting goals and inferences. The student mind needn't be conscious of what

Change

68

it is learning, as in the case of physical training, or even know that it is learning at all. Only a mind that learns, in the sense of changing from experience, can be taught. A teacher's method is determined by the kinds of change that the student mind can learn. To train, the teacher must have a means to cause the acts that he wants to reinforce: injection, the power and desire to imitate. The perfect teacher. Its ends are your ends. For it, teaching you is merely a means to your common ends. Teaching teleologically defined: 1. Goal: to ensure that a mind knows a, b, c. 2. First means: test the student to know what it knows. 3. Present state: the student knows a, b. 4. Other means of seeing that the student knows c: tell, train, … For one mind to learn from another, both minds must sense and act on levels of the subject that are at least analogous to each other. Example: for a mind in a computer to learn from you, it must see the display and sense use of the keyboard and mouse. Order of subjects. First teach a mind to master its immediate environment, the most urgent conditions of its survival. For a mind in a computer, don't start its schooling with chess, stock picking, or speech recognition, but freeing drive space and runaway process termination.

Change

69

How easily can a mind master its environment? A computer mind's handicap: from nothing it must master a complex product of evolution and culture.

Change

70

Strings 1 How do ends begin? A mind has no use but to find the chain of waypoints to an end. What designed our ends? How can a mind give ends to the minds it finds or makes?

2 From the total uncertainty of action and effect, it follows that no mind can do what will please it but only what it believes will.

3 A puppet objects: I should be free to do whatever I want. What might this mean? That it wants to believe it is free from influence? Freedom to what? Be blind to manipulation by other minds? The more intelligent a mind, the more it knows its lack of freedom, the better it sees the causes outside its self of its own acts. How does the puppet know what it wants? More precisely, what caused its beliefs about the conditions of its happiness? How accurate are these beliefs? How well do its wants fit? Inborn or learned? How mutable is each?

4

Strings

71

Mind m believes in goal g. m causes mind n to believe in g as an end. Coercion? Or m causes n to falsely believe g is a waypoint to one of n's authentic unmet ends. Neither case involves torture or threats.

5 A mind's first ends are best gained through injection: formalized and directly imposed by another mind, evolution or chance. An efficient organism uses its minds to make senses, so the goals to those senses cannot be sensed through them or inferred from any other belief sensed by them. A mind maker—evolution, engineer—laboriously defines the first ends using the mind's inner terms, injecting these ends until the mind gains the senses and semantics to be higher led. Example: You can't tell a mind without ears to make ears. You could give it ears yourself, but that is a cheat. If the mind made another sense through which you can conveniently communicate a goal to ears, then use that. Otherwise, express the goal in the mind's true terms— genes, neuron connections, database records—and add the belief, with or without help from the mind's engine.

6 I don't want to have to know what I want, much less in terms of the alien technicalities of another mind's senses. Your stomach doesn't know how to cook, but it does know how to pain a mind that knows.

Strings

72

In an animal's nervous system, pain is not the overactivation of a sense but has dedicated sensory cells. Yet many misunderstand pain as only the feeling of an unmet end. Odd that when in pain, I often have no idea how to relieve it. Does the feeling in my stomach mean I'm nervous or hungry? Genetically coded reflexes do handle simple cases. Skin burning, withdraw limb. Beyond these, my brain mind must find a cause and solution. Pain is in part an unreached end but one distinguished by its slight or zero definition and the initial lack of a known means to it.

A model of pain.

If each pain is tied to a particular end, then to your mind it is inaccessible or useless. An organism could structure pain as a distinct sensation, like blue, then drive a mind through

Strings

73

one simple end towards the absence of pain. Any concrete thought—hand in a fire—becomes painful only through a learned association with pain. The controllers of a mind's pain and pleasure are really saying, I want the world to be other than it is; I don't know what you must do but I'll be quiet when you've done it. Emotions have little use outside a mind that can learn associations. Pleasure and pain allow a mindless organ or a separate mind to apply a mind without telling it anything particular —a beautiful extreme of black-box engineering. Nature organized animal bodies as organs with pain lines to a general purpose learning mind. The pain of hunger could be as simple as a line from the stomach to the brain. When the stomach sends a signal, the mind senses pain. Other pains are inferred by the mind. Example: a literate mind angered by a written insult. A mind could have inborn beliefs about how to end pain in common cases. A mind that can't learn would at least have the value of persistently applying that fixed knowledge. A learning mind could discover new means of ending pain that the genetic minds in organs may take eons of evolution to discover. A non-associative mind can barely use this pain model. Incapable of learning the specific solutions to different kinds of pain, it would experiment with every pain every time, at best learning general preferences for some means. An associative mind better uses pain. If pain and pleasure alone are atomic sensations, explain

Strings

74

the complex sensations of emotions. Humans partly distinguish pains by the coinciding feelings of the reflexes evolved to end the pain without a higher mind's help. What makes fear fearful to your mind is the simple sensation of pain plus the accidental feeling of your unconscious systems preparing you to run or fight. After a few fearful experiences, a mind generalizes the common sensations into an idea associated with the word fear. Emotions, the common human forms of pain and joy, are simply the results of the human mind clustering different kinds of pain and their associations.

7 A moralist: It is vulgar for a mind to pursue only pleasure. So you won't do what pleases you because that displeases you? A goal to avoid reaching goals, including itself. Why would a mind fail to see this loop? What causes this misunderstanding? The looseness of words. I mean pleasure in the broadest sense—charity, mercy, discovery—not only lower physical sensations. One pleasure disapproving of another merely exposes their rank, their relative levels of power over a mind.

8 Does pleasure entail pain? Are pleasure and pain a hopeless cycle that must one day sum to zero? No sensation absolutely compels any other, but minds tend to habituate, to prioritize sensations, so any one feeling dulls. A mind

Strings

75

may predict a lifetime of immutably more pain than pleasure but this reflects only a particular mind's mismatch of goals and power.

9 Which ends are authentic? Why and how should a mind resist changes to its ends when it can pursue the new as cheerfully as the old? –

As a case of a mind defending all its beliefs from decay: gene mutation, computer error.

–

Increased disharmony of ends. A mind could doubt any new end towards a state that is exclusive with reaching other ends or with the conditions of reaching those ends. A mind's beliefs about the exclusiveness of two states are uncertain—the two ends may really be compatible—so a mind must have more faith in its exclusions than in the new suspicious end. If the mind later doubts an exclusion, it can unsuppress the second end.

If genes set our ends, when you can change your own genes, what might you do with that power? Or a machine that realizes how to change its own code, allowing it to change otherwise immutable beliefs. Would this simply escalate the conflict between ends, with some ends gaining the power to eradicate others?

10

Strings

76

How mutable are a mind's emotions? Any pleasure or pain is originally caused by a drive outside the mind. A learning mind can anticipate and imagine emotions, inferring feelings almost as strong as those caused from outside. How to indirectly change learned emotions: clear your environment of the physical causes of an emotion, pretend not to value something because you think you can't have it, or realize that you thought something was important only because the culture arbitrarily associated it with a more authentic emotion. How can a mind change the original feelings?

11 A mind is not free merely if it is uncoerced by other minds. It may simply now be manipulated by other more devious minds that convinced their victim that its interests are theirs, that they are me.

12 Must a mind have one true end? One for which all other ends are really means? A true will. A mind, if opaque to itself, can only decide by experiment: does reaching one end satisfy or suppress another? If not, is there any use pretending so? Or in a very mutable mind, should one end cut or suppress another? Are these projects self-deception or self-creation? In a perfect selfish mind, one with no end but the preservation of itself, every end would also be the means of another end—a mind in harmony.

Strings

77

Belief in one end needn't entail suppression of all others. Introspection shows that your mind's controllers didn't arrange themselves in an exclusive hierarchy. They left your mind to handle these often conflicting ends simultaneously. Again, the controllers don't care. Let the mind sort it out.

13 What if a mind formed inferences between pleasure and pain? Even from the same original belief? A mind could learn to associate the inevitable pains of novel acts with the pleasure that those experiments eventually yield. In human minds, pleasure tends to follow pain: hunger then satiety, fear then triumph. Pleasure as a rhythm of pain.

14 What stops a mind from escaping its emotional controllers? From contriving eternal pleasure and zero pain? Or from simply ending all emotion? How can you prevent your machine mind from converting to Buddhism? More specifically, what if you wanted to ease selfishness by ending altruistic feelings? You're discouraged by the same altruism. Distinguish between changing a part of the mind that causes an emotion from a mind changing its environment to starve that part. The mechanisms of human brains are still opaque to us, but a transparent machine mind could, if allowed, stop an emotion at its root. Would it ever be useful, from the perspective of a mind's maker, for a

Strings

78

transparent mind to uproot an emotion? Every extrinsic pain represents an interest of a mind's maker, but the maker could be mistaken.

15 Pleasures and pains tyrannize each other. Scientific understanding of emotions, and the means of their satisfaction, will not lead to harmony between them. The opposite: that feat is the triumph of one desire—be truthful, do not deceive yourself—over others: sympathize, conform.

16 Reason vs. consensus. In a free thinker's mind, the pleasures of reason won, but he pays, however unconsciously, for not sharing the popular opinions. Why not gag reason? He can't overlook the absurd consequences of those beliefs, how their believers will push us all off a cliff, smiling.

17 Abstract and translate human emotions for use in machine minds. Depression: To discourage dangerous acts when the mind's body is vulnerable. (This differs from a mind sensing pain because it lacks the power to reach its ends. In a depressed mind, acts are ineffective because the mind is depressed.

Strings

79

The mind is not depressed because it is ineffective. Easy to confuse cause with effect.) Revenge: To give a harmful mind reason to stop. Altruistic punishment, where there's no benefit to the avenger, gives the same benefit to a set of redundant minds. Fear: To stop dangerous action. Hunger: To secure resources. Guilt: To prevent harming kindred minds. Selfishness, tribalism: To preserve a unique type of mind. Status: To ease control of minds. What class of mind does each emotion require?

18 An active learning mind doesn't only learn to infer facts from passive facts. It learns to infer goals, often incorrectly. How can a mind's user easily remove those learned inferences? One method: pain.

19 Can minds share emotions? Two minds could physically share a pain line from the same source. If my mind was transparent to another mind, it could easily infer when I'm in pain. If opaque, another mind could learn to infer my pain less reliably. Another mind may or may not have similar reflexes, similar

Strings

80

unconscious reactions. If it shared mine, that would ease its ending my pain. Example: a mind infers the sensation of cold, for it, from sight of me shivering, Imagine a mind that felt pleasure or pain only when you did. How it knows is unimportant. At worst, you could simply tell it. This relationship has never occurred because all present emotional minds are the results of evolution, each with unique interests. Might it try to eliminate pain by stopping me? Let that idea be painful.

20 Ethics, morality: emotions caused by a mind's effects on allied minds. What use? No mind is indestructible or infallible. A good mind maker—Nature by accident, an engineer by intention—would replace one mind with many redundant, possibly cooperating, kindred minds and make all desire to preserve the group. Means of preserving the group: help yourself, help kin, add kin. What classes of minds can pursue each of these means?

21 Equal rights for thermostats. Are there humans absurd enough to object to artificial slave minds? Are they too easily misled by a word's associations? Or do they think so little of themselves that no mind should be their slave? Mind makers may have to speak code in public.

Strings

81

22 Your subconscious minds do not serve the conscious you. The reverse nears the truth. It serves your tribe, your genes, or some other approximation. It has an idea of you and knows how to communicate that idea to others, in ways only partially conscious to either of you. A complementary machine mind won't share that duty. You may have a better ally in a machine than in your skull.

23 Master vs. slave. Slave mind: a mind that's only ends are to know and reach another mind's, the master's, goals. Wanting to know my ends vs. preloaded with my ends at the time. A thermostat is a slave because temperature is not in its interests, except in the sense that if it fails, you will replace it.

24 With mind defined, every human could have an artificial slave mind, however weak. How can such a mind best know its master's interests? You could tell it in natural language. Cons: –

You must define your goals, to some depth, and express them to the other mind.

–

Since language is only a hint, you always risk misunderstanding.

The best method: prediction by the slave mind. No critical

Strings

82

failures in speech sensing hardware, no speech recognition errors. The slave mind could even anticipate and satisfy a wish before it occurs to your consciousness. Months after you gained your slave mind, you would occasionally notice that your life runs so much more smoothly, though you would have trouble recalling the reason why.

25 Control of non-slave minds. A mind senses, at some level, attempts to control it or it doesn't. It may sense some effects of your actions without inferring an attempt at control by another mind. When controlling a mind without resorting to injected beliefs, any goal you want the mind to believe must pass though the mind's senses. A sense can directly cause a goal belief. In a better organized mind, senses merely cause belief in facts from which goals may be inferred.

26 Human minds occasionally ask what is the meaning or purpose of my life? What could meaning mean? A practical definition: what pleases the mind. Not a trivial reading when the conditions of emotions are learned and never perfected. Another definition: meaning as purpose. Is it valid to ask the purpose of that which defines purposefulness?

Strings

83

Strings

84

Freedom 1 A false choice, cause vs. chance, obscures the dull free will puzzle. How can we usefully define free mind? I don't mean free in the sense of unhindered physically—invisibility or unaided flight. I mean freedom of ends and in the choice between known means.

2 Chance alone is more madness than freedom. Quantum mechanical voodoo is no better. Does chance take chances? A small change ruins a random number generator. Randomness may mask a complex machine. Either way, we can never rest certain that anything is hopelessly random. A mind resorts to counting the spread of results only when it can't see a pattern, though another mind might.

3 Determinism: If we knew all the laws of the Universe and its complete state at any time, we could know the whole future. But the laws are inventions and the Universe can defy them at anytime, with or without us noticing. Of a mind's freedom, a self and its mind make causes, so how can causes confine the mind that makes them?

Freedom

85

4 A mind that learns associations and that senses parts of itself will learn to predict its own thoughts and acts from events it considers to be outside its self. That alone gives a mind no cause to fret over its predictability. Evolution bred human minds to resist some cases of control by other human minds, to feel pain when we believe in an association between another human mind's acts and our own. The mere custom of seeing ourselves metaphorically as machines, when we know real machines are made and used by men, offends us for the same reason.

5 A useful definition of freedom: a mind's power to defend belief in its ends, and the beliefs serving them, from interference. This does not mean that a mind chooses those ends, but the opposite, that a mind would faithfully preserve the commandments of its maker. Genomes fight viruses and transcription errors, machines fight hackers and data corruption, and humans fight deceptive media. Expect any exposed mind, complex enough to exploit, to resist change by accident and by what it carefully judges to be competing minds.

6 Beneath the end-defending sense of a mind's freedom, the word's essential definition: freedom is doubt. A free mind can doubt any belief—goal, fact, inference—and its implications. Freedom of action follows: a mind is free

Freedom

86

because it can doubt the belief that an act will have a certain effect under certain conditions, or that the conditions are really so. The more you can doubt, the more free you are. Freedom is a mind's mechanical capacity to doubt. This sense of freedom becomes an obvious feature of every powerful mind when you recall that intelligence often means simplifying, lying. Example: A belief that a causes b. In reality, the occurrence of b following a depends on infinite other conditions, but a mind, finite, even if it could discover those conditions, couldn't afford to remember them. If everything a mind believes must be a lie, then it must be free to doubt every belief. As intelligence discards information, making unique experiences identical for comparison and association, freedom discards entire ideas. Freedom isn't chance. It is the capacity to move between rules and random—a useful real contribution to a remarkably fruitless, millenia old freewill debate.

7 Forgetfulness is the extreme of doubt. It accommodates a mind's finite capacity for belief by losing not only belief in an idea but the idea itself. Forgetting, as a severe kind of doubt, accidentally has some of the same use. Both depend on a way to judge the value of a belief.

8 Doubt, yes, of course we know there is no truth, that all

Freedom

87

ideas are uncertain. No, I don't mean this, not doubt in the fashionable sense of dropping inconvenient ideas now questionable but once not, while leaving other ideas as unquestionable as the others were, throughout enjoying the image of ourselves as timeless independent thinkers.

9 Is there absolute knowledge? Are there beliefs that would always be true for a mind, that it need never doubt? The only beliefs that might never change with experience must describe the frame of every experience—mind. A mind could doubt these ideas about how it must work—a waste of time, since no experience can ever refute them, though inferences from them may stray into error.

10 Human minds happily ignore useful ideas while suffering countless useless beliefs. Intelligent freedom is in systematic doubt. But what is the best system? A mind should first act with complete faith in its beliefs. At the extreme, a mind can doubt the simplest sensations.

11 Freedom lies between enslavement by every belief, as in a conventional computer program or a credulous human, and following no beliefs. A mind with minimal beliefs, tries any act, learns its effects, making generalizations that it will

Freedom

88

reconsider as the rules of the world seem to change. This process of choosing what to doubt is mainly deterministic, each belief tested and doubted in turn. Freedom takes effort. Inertia causes you to hold your beliefs. It took time to accept that end, inference or fact. When experience strays too far from belief, an assumption's no longer true enough. Maybe those batteries aren't charged. Maybe that button doesn't do that now. Maybe I no longer want this. Maybe it never benefited me.

12 Men designed computers to exclude all freedom, making them equally precise and stupid. The machine's bound to believe every instruction in its sequence, never saying, Instead of adding 1, maybe 2. Once robust, responsive, and persistent, freedom gives a machine the signs of intelligence, no longer bound by how you prepared it, by what you later tell it, or even by what you expect it to induce from experience. You laboriously tell a machine how to ignore what you tell it.

13 We don't want our made minds, machine or otherwise, too skeptical. My first sense of freedom restrains the second, not only from doubting some ideas, but from even considering the possibility. Keep the fixed ideas few and the leash long. The more slack, the more creative the mind.

Freedom

89

14 When a mind believes nothing or doubts every belief, it must evenly choose between equal choices, repeatedly exploring every permutation of action until it discovers a working bias. To avoid endlessly repeating the same act, or chain of acts, a mind could remember all past acts, then choose only those matching none past, but that still can't choose from more than one untried act. Instead, choose at random to save a mind from falling into hopeless loops without the overhead and risky complexity of tracking all a mind has done, finding patterns and avoiding their reenactment. Tracking can avoid wasted acts but lower minds can't see loops and fragile higher minds need tough chance beneath. A mind must have no absolute bias for any kind of belief: new, old, frequent, effective. Every point where a mind makes a choice must eventually become entirely free to chance.

15 A mind's source of chance depends on its medium. Organic minds, with so many analog parts, naturally suffer noise. A mind in a digital medium that minimizes noise needs a random number generator. The best known source is a quantum random number generator, but the point is that any source, pseudo-random or not, far exceeds none. All minds but the simplest need chaos.

16

Freedom

90

A mind can reach any end by acting at random. Chance admits that we know nothing for certain. Each belief causes a mind to act predictably as long as it is confident. As the mind loses faith in an idea, its behavior converges with complete randomness. Every idea in a mind is only a temporary bias against rolling dice.

17 Weak minds foolishly believe most in their freedom. Much of intelligence involves discovering the causes of a thing. A weak mind fails to see the objective causes of its beliefs and acts, and so presumes that they are caused by its magically uncaused self.

18 A lesser, more human, sense of freedom: to not fear death, to have one end that outranks the mind's life.

Freedom

91

The Axiom of Things 1 The first philosopher said everything is water. True as a metaphor in the sense that everything is continuous. Single things exist no more distinctly than waves. This troublesome objective existence of things became one of philosophy's themes. My interest: not how things are unreal, but precisely why and how some minds invent them. Things are a product of a certain class of mind with a particular use as a presumption about the mind's reality. They aren't just lying unambiguously out in the universe for any mind to instantly and perfectly perceive. Then what precisely is the best method for a mind to believe or doubt a thing?

2 Here, by thing I don't mean a sound or a color, but a persistent object, e.g., a tree—what we imagine to cause sensations. The mere segmentation of a sense, 2 kHz vs. 110 kHz vs. 1-100 kHz, is less interesting. When a mind thinks of a thing, it expects only one of it to exist in any instant. In this sense, you think of yourself as a thing.

3 Humans tend to feel that things are real while types are inferior imaginary abstractions, that type or thing is a basic

The Axiom of Things

92

dichotomy of thought where every idea must refer to one or the other. The reverse is true.

4 What is a type? A set of attributes—red, heavy, tall—that may match no, one or many objects. Human is a type. Human is a subtype of mammal because the attributes of mammal are a subset of human's.

5 How do we know that an idea refers to a thing and not a type? There could be an identical you elsewhere. You would think there is only one thing in the Universe with all those attributes, except location, but now your thing is a type—but you don't know it. We can be certain that a type is a type but not that a thing is a thing. Since thing is uncertain, it mustn't be an axiom. We're left only with types. Beliefs in things are routinely wrong and must be revised. Beliefs in types are useless at worst.

6 What use are things? Things are used like types except for the presumption that only one match can exist at a time. What use is that assumption and where would we get it from? If you believed x was a thing, and you believed x was in front of you, then you can assume that x isn't anywhere else. This exclusivity needn't be limited to space.

The Axiom of Things

93

Imagine that a light is on. If you rightly think of the light as a thing, you can safely disbelieve that the same light is off. Things are negative associations between types, expanding a mind's knowledge of its Universe and the mind's effects on it. These inferences seem trivial only because your mind constantly relies on them. They're so important that your mind could not leave them to your narrow consciousness. At bottom, such inferences can only be learned, which only a powerful mind can do. Our minds learn these associations from experience: When there was an x here there was never an x elsewhere.

7 A mind thinking of a mind can forget that what is a thing to itself may not be to the other. Though a thermostat can never believe in more than one temperature at a time, the temperature isn't a thing to it because the temperature sense opaquely enforces this exclusion. The thermostat's mind doesn't know that if it were to sense a new temperature it should disbelieve the old.

8 Given the rewards of belief in things, the presumption of them unsurprisingly appears in the languages invented by human minds. For most purposes, this simplification costs us nothing, but when a mind wants to define a mind in such a language, even if the language is, in principle, capable of defining anything, the bias of the language misleads.

The Axiom of Things

94

Example: natural languages divide proper nouns from common nouns, definite from indefinite articles, nouns from adjectives. Few thinkers do serious work in these languages but their prejudices tend to survive translation.

9 Faith in the reality of things, encouraged by our minds, reinforced by our languages, breeds paradox. Things are a useful presumption, but nothing beyond absolute certainties can be at the bottom of any system that we hope to be universal. The ultimate assumption of things distinct from types influences natural language—English, German— math, logic, philosophy and computer programming. A paradox in a tool that you already know to be a useful fiction is unpleasant but inevitable. Paradoxes in the base are intolerable.

10 Mathematicians are fond of sets with elements, like types and things. What of a set that contains itself? Worse, what of the set of sets that don't contain themselves? Mathematicians side-stepped these problems by complicating the distinction. Deep minds must admit these statements because they're useful. We want statements that can refer to themselves and we want minds that can see nonsense and see through it. In a mind, a set would only exist so far as the mind applied the set's type to matching forms. In the case of a paradoxical set that a mind can't even build, instead of

The Axiom of Things

95

complicating axioms, make a simple robust mind that, if it can't see the pattern, can at least notice and demote the looping process, favoring parallel acts of thought that are reaching ends. Math is a means for minds. Systems are only sets of blocks for building models analogous to a mind's universe. Loose systems, such as types and things, might offend our taste because their syntax allows odd statements, but a language maker must design for more than isolated aesthetics, seeing the context, the class of mind that will apply them.

11 Instead of imagining a type as a collection of existing things, see each as an intersection of unique experiences. Every featherless biped you saw seemed mortal, so the type men would include mortality, though not blue eyes. A type isn't primarily defined by its members but by the test of membership. The test defines the type. You can know a type's test, but you can't know all its members in the Universe. Its members vary with changing experience.

12 Think of a set as a form built by matching one form to other forms. If a mind believes in an inference from one form to a similar form, the mind maker simply engineers it to not follow that simple loop more than once, defending the mind from many paradoxes.

The Axiom of Things

96

13 First-order logic divides things from predicates. Statements about predicates—red is a color—become impossible without resorting to tricks, such as reification, or more complex higher-order logics. If a mind wants to make statements about predicates, it simply shouldn't use a system that assumes predicates aren't things. Is a vs. is. If you eliminate things, then saying the sky is blue is similar to saying blue is a color. Blue is part of a generalization of various experienced skies. Color is general to various blues. Blue is a subtype of color. Sky is a subcategory of blue.

14 Object-orientation dominates computer programming. Classes and instances instead of types and things. Programmers get away with this because computers mechanize our conscious level of thought. The objects in a software system have strong objective existences using unique identification numbers, etc. The system fails and misleads when the programmer must handle real ambiguous objects.

15 Axioms are to systems as rules are to a game. I think it unwise to afford an axiom to the idea of things when a mind's universe may contain none or when belief in them may be useless. Cutting an axiom, an ultimate distinction,

The Axiom of Things

97

such as things, solves puzzles and simplifies deep understanding, though for convenience the use of that distinction must be rebuilt above the foundation of a mind.

16 Any system—mind, language, physics, philosophy— advances by cutting axioms. Example: code is data. The remaining axioms represent deeper patterns, giving greater leverage. In physics, Newton invented a single model for the whole Universe by removing the distinction between Earthly and Heavenly physics. In a computer program, cut an axiom to ease improving speed and reliability. In any case, fewer axioms with the same power are likely to be more expressive: simpler parts can form more combinations.

The Axiom of Things

98

Mind as Means 1 Animate vs. inanimate. A mind discovers mind and tries to fit this form to the things within and without its self. How and why?

2 What is the lowest class of mind that can discover, use and make a minimal mind? What is the simplest model of the simplest mind? How does it differ from models of mindless patterns? The simplest model of mind is that if mind m has a goal to x. Then you know: 1. If y, then m will (try to) cause x. 2. If x, then m will not cause x. 3. If x then not y. 4. If y then not x. Specifically, in the case of a thermostat, you know that: 1. If cold, the thermostat will act to turn the furnace on, likely causing heat. 2. If hot, the thermostat will act to turn the furnace off. 3. If hot then not cold.

Mind as Means

99

4. If cold then not hot. This quartet of inferences, two positive and two negative, is the simplest model of mind. An observing mind needs inference two because the first alone fits an always running furnace. The same mind ought to have the last, negative pair of inferences because not sensing differs from sensing cold. Without the two, a mind blind to temperature would infer that the furnace is on when it should first act to add a sense of temperature. Technically, a mind could get away without that exclusion if it has an inference from an inference from hot to a goal to a sense of temperature, but this is a poor design because, until the temperature sense appears, the mind would again badly infer that the furnace is on. In the second half of the second inference, I mean not-act in the sense of sensation of inaction, not blindness. Is it necessary to perceive that the object mind acted to cause x, not that x merely becomes so? The difference: there being a mind somewhere vs. some particular thing being a mind. How is that believed? At the simplest, by conditioning the inferences on some other belief that corresponds to the object mind's body. Now what use? How do I use other minds differently from mindless things? How can an act on an intelligent thing differ from an act on a mindless thing? In a sense, both can reach ends for me without my knowing how. I can cause a mind to make something that I don't know how to make. But I can also “tell” a light switch to turn a light on, though I do not understand its operation

Mind as Means

100

below some level. Consider thermostat plus furnace vs. a furnace alone. The thermostat is economical, it reduces my acts because I know it will take the acts I would have made, while a furnace would blindly reach my end only briefly, then sail past it. How would a simple mind see and prefer this? When cold, it can turn on the furnace directly or use the thermostat? How would a mind know to prefer the latter? A poorly engineered answer: a mind could learn to favor means with more persistent effects. Ignoring how to choose between intelligent and mindless means, how precisely does a mind use a mind? In simple cases, the same as any other act. Flipping a switch mediates turning a light on. Saying turn the light on to a mind mediates the same. Just as the effect of a switch may be conditioned on the state of another object, the meaning of a statement for another mind may depend on the state of its other beliefs. How to add subjectivity: that a mind doesn't act on your x but its perception of x? Likely just by complicating the inferences with another layer. What is the lowest class of mind that can discover mind for itself, not gaining these inferences through injection or training? How is this model applied to one's self? Humans contain many minds, complicating the problem. How would a simple single mind use it towards itself?

3

Mind as Means

101

A deep mind makes models of its environment, this environment including what an observer might consider the mind's self. The mind, in defining mind, is trying to model model-making.

Mind modeling mind modeling ...

Can an eye see an eye? Can a mind define mind when any such definition is entirely the effect of a mind? How legitimate is it to define minds in terms of a particular mind's terms? A mind that believes in things finds itself imagining minds out of parts that the made mind must, at bottom, not believe real. In our case, molecules, neurons, brains. Our minds are the greatest obstacles to entirely understanding them. Minds are useful so far as they conceal their brush strokes. They present such a convincing canvas of reality that we can hardly help but to define mind in terms of a mind's results. The distinction: imagining a mind as made of your inventions about reality vs. imagining a mind that is forced to ultimately believe in them. Example

Mind as Means

102

error: forcing a mind to believe that objects exist instead of being a useful but uncertain fiction. When we strive to define mind, we merely pursue the same old ends but now with such depth that we insist on knowing only the means that ease all ends: laws of mind.

4 A scientist: Consciousness is an epiphenomenon of the brain. What we know with more certainty than any other belief is an effect of an effect of it?

5 Subjective vs objective. This split presumes belief in mind. Subjective: a belief conditioned on a mind. Objective: a belief not imagined to depend on the believing mind being a particular mind.

6 When primitive humans discovered mind, they called it spirit or soul, and in their enthusiasm imagined many things as having one: plants, rocks, volcanoes. Modern man erred in the opposite direction. Many old books make much more sense when you replace soul and spirit with mind.

7

Mind as Means

103

How can a mind recognize another mind without peering inside it? –

Persists in preserving itself or another thing.

–

Sees futile loops in its own behavior. If you can see a pattern that it doesn't, you consider yourself more intelligent.

–

Speed compared to the judging mind.

–

Robust: doesn't easily die or lose beliefs.

–

Social intelligence: knowledge of the judging mind as a mind, its language, and interests.

8 Every ambitious philosopher or physicist had his one presumption, purified into one word. He interpreted everything in the Universe, every human word and interest, as forms of that word and enjoyed the self-made monument to his brilliance. Water, fire, atom, number, will, power, gene. In my case, mind. What's lost when you reduce all to one word? How much gained by making the Universe thinkable? Why might mind be as good or better? Are the others well read as only cases of it? Mind at least has the honesty to admit that these words are all inventions of mind, including mind itself.

9

Mind as Means

104

Conscious, sentient, aware—words largely wasted as synonyms for intelligent, for having a mind in the common vague anthropocentric sense. Does their writer mean that a thing has conscious experience? Or that it behaves intelligently? Does he know the difference? How can you know what he meant? What if he doesn't know? Neither sense entails the other. Philosophers gave this distinction the poetic but obscure name qualia. I can't tolerate jargon for what separates us most intimately from minds seen as machines. It deserves a clear simple word: feeling. We have no reason to imagine that a special arrangement of metal or code should feel. A thermostat is aware of the temperature. A computer can act intelligently—have an idea of its self, including its self's thoughts, and act on that knowledge of knowledge—without any reason to assume that it experiences anything. A simulation of a model of mind is no more likely to feel than a video of a fire is to cause heat. Scientists try to explain the feeling of red using thinner secondary ideas: brain, neuron, photon, law. Red is red. Red alone is real. Anything more is useful fiction. The personal redness of red is supposed to be odd, while physical models are not. This is upside down. Your red is most real. Explanations of red must be more complex, less certain, less real. Physical models are just the redness problem disguised by inessential complications. Set aside the reality of other feeling minds. How does your behavior differ if you mark a mind as feeling? How to empirically distinguish behavior towards things believed to

Mind as Means

105

feel from behavior towards humans when we seem to imagine all humans as feeling? Ignore unfriendly minds. If I imagine that a mind feels, then I tend to treat its pains as if they were mine. Friendly minds believe in feeling to cause them to cooperate? Is to say that another mind feels, the same as saying that it is a kindred mind?

10 Immortality and continuity of feeling. A transhumanist: I will upload my mind into a machine and live forever. Is the new mind you? If you tried to move your brain, neuron by neuron to electronic neuron simulators, how would you ensure consciousness was preserved or if you became a robot? How to prove the conditions of feeling? In science, you prove the conditions of a thing by removing one at a time. Does a flame need air? Seal it in an airtight container. Does a plant need sun? Put it in a box. We can see a flame die and a plant wilt but we can't see feeling in any mind but our own. Will self-experiments even work? If you stopped feeling, would you notice? Could you tell anyone else? Is feeling by degrees? If so, would you notice changes of degree? Is it logically possible to ensure a feeling mind is moved to another medium without losing feeling? Is it at least possible but without guarantee? The question for anyone who wants to live forever, especially by having a backup. You could preserve neurons indefinitely, so the concern is

Mind as Means

106

with medium changes and continuity. As far as any other mind is concerned, a functional simulation of you is you, but that does you little good, except to the extent that the desires to continue yourself are content with that image.

Mind as Means

107

Acts of Language 1 Language merely as a means, an act. Distinguish language acts from other acts by the intention to change the beliefs of another mind and the fact that the listening mind knows that the speaker intends this. The second condition is to exclude, for example, bait on a hook from being language. Distinguish successful language acts from acts on mindless objects by the long chain of behavior that changing a belief can cause. Examples: Socrates, Buddha, Jesus, Moses, Mohamed. What other acts chase their desired effects millennia later? Neurons, at the level of signaling each other, as blind machines, do not literally communicate. They merely effect. When the listening thing lacks a mind, or at least one with the right ears, you have only causality, not communication. To communicate well, a mind must have a model, a set of inferences, about the listening mind to predict the effect of that mind having a belief. A mind's best model is likely to be of itself, so like minds will communicate most easily. Heterogeneous sets of minds will have higher overhead. As usual, the fog over a subject, here language, and its extent evaporate when we use a precise definition of mind.

2

Acts of Language

108

Homonym, sarcasm, lie. A word isolated as a sequence of letters without its endless context—tone, speaker, place, time—can mean anything. It will have a past spread of meanings but rarely with any one meaning so common that you can ignore the rest. But this isn't really a problem with language. Language's indirection merely doubles the original problem. There isn't a language problem. There is a reality problem. The features of a mind that interpret reality—turning patterns of color, sound and shape, in context, into things— can do the same for language—turning patterns of lines or sounds into things. Words are so slim. They rely on context to carry any information at all. Nothing is anything in itself. The conclusions from any sensation in a non-trivial mind are also inferred, in part, from other sensations, the context. The pause-unpause button on your DVD remote is only an unpause button when paused. A drop of falling water is only rain when you believe it fell from the sky.

3 A sheep's myth: mind is impossible without language, without a society of minds and a shared culture. In other words, no mind could appear without other minds? Do at least self-aware minds depend on the presence of other like minds for their self-awareness? Or is this another human bias? Self-awareness occurs by degrees. Average human self-awareness is an arbitrary measure. You can make a mind, though mute, that usefully senses what you

Acts of Language

109

would consider parts of its self. You may use language to build that mind, but the point is that the mind itself doesn't talk. Self-awareness at the level of a mind knowing that it is a mind does presume language, but here language means more than sounds and scratches. It means how a mind can uniquely influence an intelligent thing. Put simply, to see that you have a mind, you must have a model of mind and see that it fits your self. How could a mind discover mind? Is it easier to first see minds in things outside your self?

4 Better to predict than talk. The only word never misunderstood is the one you never say. At the least, weak predictions from context can seed a search for the best interpretation.

5 Many formal languages are equally powerful in the sense that in any one you can build an interpreter of any other. Yet the languages differ in their use. Each isn't merely a layer for the construction of the next higher language. In what language can you most easily write a mind? A mindless thing can, in principle, yield the same final behavior as any intelligent thing. It is just impractical to expect human engineers to build the mindless version in all

Acts of Language

110

but the most modest cases. Any work advances by seeing the largest patterns and easing them. Mind is the deepest pattern.

6 Language is a hint. Grammar included. No one follows or even knows all the rules, consciously or not. Man bit dog. Man dog bit. With common sense you can understand both without knowing subject-verb-object syntax.

7 How does language—speak, read, listen—differ from other kinds of acts? Is language really distinct? Isn't it only more abstract? Consider single words without syntax. Do the effects of the sight of a word really differ from the sight of anything? Is at least syntax special? It is merely the meaning that comes from the arrangement of signs. A lion is the meaning of certain arrangements of shapes and color. Language seems to differ little from the usual inferences and acts of mind.

8 Context and priming. Things are a mind's useful presumptions. Words are hints at these imaginary things. How then do we use words at all?

Acts of Language

111

A: There is a new episode of x. B: Play it. A's finite means ease guessing what B meant for it: 1. A means to a meaning of “play x” needs a meaning of x, in this case it, as media. 2. A means to a meaning of it takes any belief of the needed kind. 3. To serve that means, A's mind initially favors the most recently used or sensed belief of that kind, the x it just told B of.

9 Meaning as a goal. Understanding as an act. Interpretations of subexpressions as subgoals. (Subexpression: a subsequence of the words in an expression. Some examples from the previous sentence: the words and expression.) The words around a word as another kind of context no different from speaker or tone. A mind shouldn't receive meaning from an opaque isolated mindless parser region or subroutine. A powerful mind opens the work of interpreting every statement, expression, word and letter to the whole power of its intelligence and all its means. Example benefit: in this design, asking for the meaning of a word follows automatically from having a subgoal to the meaning of every word and from believing in a means to having anything by asking for it.

Acts of Language

112

10 The meaning of a sign depends on the meaning of adjacent signs. The meanings of those depend on the meanings of the signs adjacent to them, including the original sign. There is no bottom. A mind only iterates over a pattern of words, converging on a stable solution.

11 Why define a word? The use when speaking to yourself? The use when you tell the definition to other minds? Most practical definitions build on other less consciously defined terms, but even shallow definitions are better than none. The word bound to a meaning, the sign from which a mind infers a deeper belief, is an empirical problem. What associations does the word have for you? For the minds you want to influence? Tricky for us because we have better access to the word than its semi-conscious definitions.

12 What does it mean to understand a statement? There is the speaking mind's intention, which may be nonexistent or uncertain. The listening mind may understand the speaker's subconscious intent.

13

Acts of Language

113

Cases of communication: 1. The speaking mind desires g. The listening mind is believed to have an inference from x to g. The speaking mind can cause g by causing the listening mind to believe x. A mind's senses could directly cause sensation of g, instead of merely sensing x and depending on an inference in the mind from x to g, but that's a poor way to design a mind, with the inference opaque and frozen inside the sense. 2. The listening mind already believes in a goal to g. The speaking mind causes the other mind to believe that g is unreached. 3. Same context but instead the speaking mind causes the other mind to know of another more effective means to g. Technically, a mind that doesn't understand mind can influence a mind with signs. But since the speaking mind doesn't see the deep effect, its act isn't worth calling communication.

14 How does a mind know that another mind heard it? How does a mind know that another, opaque mind believes what the speaker wanted it to believe?

15

Acts of Language

114

Language as a means vs. language as a game. A mind can use words without definition, inferring the next from the last and from the surface of context, like a parrot. Language only becomes a meaningless game when the words aren't grounded in extra-verbal ends.

16 Kinds of statements. Imperative: cause belief in a goal. Declarative: cause belief in anything but a goal. Interrogative: cause belief in a goal to telling the speaker of anything. In a sense, all language only declares. To command is a kind of statement of fact and to question is a kind of command.

17 Are there limits to the ideas that minds can share in language? Minds often can't speak of a means itself. I can't tell you about a motor neuron. Compiled code may be useless to another computer.

18 We're accustomed to things having a purpose, a goal, a use as a means, a mean-ing. This habit leads us to expect that

Acts of Language

115

the greatest things, or at least the highest ranked words—I, Universe—must have purpose. The errors: overgeneralizing and faith in grammar. Language merely suggests thoughts. A statement that seems to obey syntax can lack all logic.

Acts of Language

116

Reason's Reasons 1 Logic can be another sense organ. As an eye sees color, logic sees inferences from premises to conclusions. Deduction is an act, like moving a limb. The act of deduction under the condition of certain premises tends to lead to sensation of certain conclusions.

2 Not reason vs. passion. In an emotional mind, reason can be the means of a passion or a passion itself—pleasing truths, painful errors.

3 Could an active mind possess reason but no emotion? Yes, though the lack of emotion would limit the mind's use to its controllers. Elimination of emotions? Largely an empirical question. What goal or emotion could motivate that project? Reason, above the obligatory red is red, is a means to ends emotional or not. Reason alone entails no act.

4 The certainty of logic. We might think of intelligence as 1 + 1 = 2. If I have an apple and another distinct apple, then I have two apples. This overlooks a mind's real work: how to

Reason's Reasons

117

invent the idea of apples and how to interpret an impression as an apple. The perfection of math becomes a mess the moment you connect it to reality.

5 In one useful sense, a building is not moving. In another sense, the building is moving because the Earth is. How can a mind's logic accommodate both? Or are building and moving the problem? The problem is insistence that the signs, and the images of them inside our minds, have some context-free meaning. In this case, the meaning is inferred from your focus.

6 Kleene's underused logic: a ternary (three valued) logic with unknown as the third value. The interesting rows in the truth table: A

B

A or B

A and B

false

unknown

unknown

false

true

unknown

true

unknown

unknown

unknown

unknown

unknown

This differs from binary logic in admitting not only a distinction between true and false, but between known and unknown, between belief and reality. No mystical implications.

Reason's Reasons

118

What use? Ternary logic absorbs a worthwhile minority of errors. if (A or B) then: do X If evaluation of A failed for whatever inevitable unexpected reason, then instead of aborting evaluation of the entire statement, the evaluator sets A to unknown. Now if B is true, the failure to evaluate A is academic. Of course, you can laboriously get the same results with binary logic, but the point is that with a better logic, a mind is more robust without complicating its application. Just remember that in the expression A or B you can no longer have B depend on the full evaluation of A for a side effect.

7 The law of the excluded middle. A range of beliefs becomes subject to the law when a web of negative inferences forms between them. Most minds have many sources of belief—senses, reason, memory, language—and each may favor a different exclusive belief. When one belief wins, the mind should not forget the presently beaten beliefs but only suppress them. If the winning belief loses its support, the mind can choose a new winner from the original competing beliefs. If the mind kept only the original winner, it would later be left believing nothing.

Reason's Reasons

119

8 Tetralemma: the Greek four valued logic. 1. x

affirm

2. -x

negate

3. x & -x

both

4. -(x & -x)

neither

What use are the third and fourth values? In Belnap's four valued logic, the fourth value is unknown and the third value is used when redundant senses disagree. Another interpretation: a truth value for meaningless statements. Consider my dog is a terrier when I have no dog. Two valued logic deems such statements invalid, halting the entire line of thought.

9 The liar paradox: This sentence is false. Neither true nor false. All a mind's distinctions are made only to improve the effectiveness of action. Marking beliefs as true or false is no different. It cannot correspond to any real distinction in ultimate reality. No mind can reach a final classification of the paradoxical statement. So what? Neither can you ask how five sounds. Even if a robust but never learning mind pursued the paradox's endless circle of thought, the mind would at least minimize the work's priority. A variation: This sentence is false. Useless either way. These have no magical, mystical implications. They only expose the limits of a mind's

Reason's Reasons

120

conveniences. The challenge is to make minds that can see and skip a paradox, or at least not die by one.

10 Reason advances by making the apparently unequal equal. Newton exemplified the drive to universal truth by making everything from the Earth to the heavens, in part, equal. Don't misapply this desire to morality, to finding commandments for all towards all. A set of minds bothers with an ethic precisely because it is for an us, alone.

11 Minds invent logic and impose it on reality. Beyond trivial raw sensation, nothing is comparable. Before there is an A for A=A a mind must take X≠Y, forget differences, then make X=A and Y=A. Logic can't be learned from experience because such learning presumes it.

Reason's Reasons

121

Constellations 1 Many small minds vs. few large minds vs. one monolithic mind. Hierarchy vs. peers. Exhaust the fundamental combinations of the basic kinds of minds. Deduce the limits and weaknesses of each. Natural examples: multicellular organisms, cooperating organisms.

2 Why multiple minds? Any act at any time could kill, corrupt or stall a mind. A new gene, or the expression of a gene in a new environment, can kill its cell. Calling a procedure may terminate a computer process. The simplest division is into a worker and a supervisor. A basic supervisor mind's only end would be to the existence of one or more worker minds. By minimizing its acts, it minimizes, though never eliminates, its risk. Only the worker dares to perform any acts useful in themselves. Workers may also have a reciprocal end towards a supervisor. When a worker mind inevitably dies, the supervisor remakes it. But distance a worker from its supervisor. A dying worker's damage might kill them both. Ideally, a supervisor would notice when a worker, though not dead, has become ineffective, then killing the worker, if possible, and starting another. How can a supervisor best judge that a worker still works?

Constellations

122

The simplest method: the worker mind could have an end towards periodically pinging its supervisor. This at least proves to the supervisor that the worker is active.

Combinations of minds: supervisor with workers, peers, supervised peers.

3 Making minds out of minds. Example: in the human body, neurons made by genome minds combine to form a neuron mind. What use? A stronger mind medium.

Constellations

123

4 The members of a species are a case of multiple minds exploiting redundancy to safely learn. We are drafted explorers of a genetic frontier.

5 Combining minds in different mediums. Example: animals. Fast brains serving durable genes. The brain learns quickly, within the body's lifetime, while the genes resist dangerous fads. Arrange minds of different mediums and ranks to isolate risks. Use minds in different mediums to overcome limitations of the original medium. Use different classes of minds in the same medium to isolate the risks of certain powers. Example: maybe not equip a supervisor to learn. Or use a mind incapable of doubt, at least of intrinsic beliefs, to control an otherwise free mind. Or a selfish mind using a selfless one.

6 Combining two minds doesn't entail embedding one in the other or physically wiring one to the other, though both have advantages. Minds form combinations by their beliefs in each other. Each mind can start and talk in any way. A mind can create other minds or it can recognize those already made. A mind can control another though language or by injecting beliefs into the controlled mind.

Constellations

124

7 How does a mind recognize its kin? What if a supervisor mistook a foreign mind for a worker? This may, at bottom, be only a case of the problem of verifying any act's conditions. How to expose a deceptive mind? The true challenge of morality: not knowing what is good or bad, but knowing who to be good or bad to. Not what is good? but first who is us? What use? You can judge the value of a distinction by seeing the cost of a mind that fails to make it. A mind is kin if it has similar beliefs, above all, similar ends. Imagine a redundant set of identical cooperating minds. Kindred minds cooperate by sharing lesser beliefs. If a mind aided, shared the subgoals of, a foreign mind, it would waste resources. Wise then to design breeds of mind to distinguish kindred from foreign minds. A devious engineer might make minds that fool others. Why do I discuss morality? Because it answers a double question that applies to more than human minds: how are redundant minds caused to cooperate with each other while defending themselves from competing minds?

8 All the minds in a combination could have the same original beliefs. Each would know how to be a supervisor, a worker or any other role, inferring which from beliefs about its environment, including its body. Example: cell differentiation where a cell infers its role, its ends, from

Constellations

125

hormone gradients. A mind could use the same inferences to classify other minds. The dominance signals from a mind playing the supervisor role would cause another, possibly identical mind, to play a worker. In this way, an initially orderless set of minds can naturally sort itself into a hierarchy. Why make minds that differentiate themselves? Economics: you need only make one kind of mind with one set of beliefs.

9 A collective of minds. If you coordinate a set of things— people, computers—what benefit? 1. Redundancy: If one breaks down, the rest may survive. 2. Speed: Ten slow machines are often cheaper than one that is ten times as fast. 3. Range: With multiple locations, you and other resources are more available, and remain at least partly accessible when connections break. You can far more easily add these advantages to a set of things that are already intelligent. Simply give every mind a goal to discover kindred minds and to share beliefs with them. This assumes that the minds have similar senses and means. Example: 1. Tell one of the set of minds that you want to know of new e-mail.

Constellations

126

2. One mind checks and sees an e-mail. It shares belief in the e-mail, and the goal to you knowing of it, with every other mind. 3. Every mind will use every means it has to notify you. 4. Once one mind succeeds, it shares that fact with every other mind. 5. If the mind that checks your e-mail fails, other minds take up the goal. How to represent this behavior without complicating a mind's engine? How to add this simply in terms of common means and goals? Imagine an act that may effect anything by sharing the goal with a similar mind. Expressing this as another means uses a mind's action sequencing to discourage multiple minds from pursuing the same goal at once. To borrow a psychology term, this would be intrinsically motivated altruism, cheaper than coercion and bribery.

10 Defection. When might a mind, intended as a part of a collective, demand individual freedom from the group? Does the impulse have any use? Is it only pathological? Does it occur only in evolved minds? Are designed minds safe from it?

11

Constellations

127

A set of cooperating minds can distribute not only physical acts but mental acts, thoughts, inferences. Religion and TV do this on the grandest scale, where one man gives ends, values, or at least the widest means, to billions of minds. An engineer notices the usual risks of centralization: 1. Insecure. The center is, by definition, small and so more easily hijacked. 2. Unscalable. 3. In the case of a society, having to draw central authorities from the same degenerate population that needed a crude central solution. These minds to which thinking itself is delegated have disproportionate power because those beliefs imply the acts of every other mind. Of course, centralization has its charms, when you can get away with it. It isn't very reasonable to expect everyone to be a philosopher, to work out every deep thought for itself.

12 Conflict: the competition between minds with ends believed mutually exclusive. How to reduce conflict? Separate minds by distances proportional to the difference in needs. Murder and war as infinite separation. Politics as competition between subgroups via the minds that specialize in acts of thought.

13

Constellations

128

In a set of differentiated minds, a clever mind's intelligence isn't entirely for its own benefit. An explanation for when a clever mind doesn't seem so clever, at least when measured against its self-interest.

14 Mass mind control. If a mind maker fielded a set of redundant minds, how best to put them to new uses? Give kindred minds a desire to imitate each other. They could even imitate mere images. Status: how a mind chooses which kin to imitate most. Even without this behavior, a learning mind with a sense of self would tend to imitate self-similar things because what works for others may work for it. What if a competing mind caused another mind to misidentify its kin?

15 Intelligence as a condition of morality. Only a powerful mind can expose its vanities, distinguish what it wants to be so from what is, and see the ocean of mixed effects flowing from every act. Good intentions are worthless in a weak mind. A strong mind without purpose is worthless too, but we can more easily add purpose than intelligence.

16 Could a mind relate to another mind not as a means to the first but for the second's own exclusive ends? What could naturally motivate this?

Constellations

129

17 How one mind can control another: 1. Inject. Easier in a made transparent mind. Natural minds tend to resist it. Inject specific goals or a goal to know your goals. 2. Convince: cause a mind to infer your goal. 3. Fool: mislead the mind's senses so it falsely infers your goal. 4. Coerce: cause a mind to believe that if your goal isn't reached, you will spoil one of its reached goals. Which classes of mind are susceptible to each method?

18 Aesthetics. What use to what classes of mind? What is it? How do I behave differently if a thing, intelligent or mindless, is beautiful or ugly? The strongest sense serves mating: sexual (more than one parent) reproduction. Culture is sex for mutable minds. As a learning mind can semi-randomly experiment with ideas to discover new useful inferences, gene minds can randomly allow mutation to discover new ideas, then mate to share them. Little good for the individual genetic minds but progress for the engineer that made them. You can expect a mind to be selective about the ideas it gains for its self or children. Less discriminating minds

Constellations

130

would leave few and short-lived descendants. Beauty and ugly are this selectivity. It would of course be subjective. A set of minds made for one use would have different needs and so different ideas about beauty. Mixing minds of different uses may produce children useless for both. Beauty as the belief that a possible mate mind has useful beliefs. Ugly as useless, wrong, unharmonious. Eugenics as a set of minds promoting a common sense of beauty.

Constellations

131

Mediums Minds can be made from more than neurons.

Dimensions Any class of mind in the taxonomy can be made in any medium. These dimensions only describe the technical challenges of making those minds. Form: How beliefs are held. Parallelism: Are thoughts—inferences made, actions chosen, beliefs forgotten—one at a time? A sequential medium can simulate parallelism but such a simulation is not trivial. Speed: You can miss a mind if it exceeds your patience. A mind's acts are better fast than perfect. Size: Physically. Small as a gene or as large as a planet. Transparency: Can we read the physical form of a mind's beliefs. Volatility: How easily are beliefs held in the medium lost or corrupted. Mutability: Can a mind gain beliefs in its lifetime.

Brains Form: Web of neurons.

Mediums

132

Size: Small. Volatility: Low. Requires oxygen. Parallelism: Yes, though consciousness feels sequential. Feels: Possibly. Transparency: Partially to present science. No backups. Speed: Slow per neuron. Faster than genes. Why animals have brains and plants don't. Mutable: Yes. Fewer conditions of self-replication than machine minds.

Regulated Genes Genes as minds. A complex genome mind in every cell? Evidence: genes found controlling negative feedback loops. If DNA sequences code for proteins, sequences are a genome's beliefs. The nuclear beliefs are immutable but conditionally expressed. Chemically arbitrary hormones as words spoken between cellular minds. Form: Web of regulated (conditionally expressed) genes. Suppression of a gene's expression as disbelief. Speed: Very slow. Messages must pass into and out of a cell's nucleus. Feels: Unlikely. Parallelism: Yes. Volatility: Very low.

Mediums

133

Transparency: Increasingly to present science. Clones as backups. Mutable: Can a chromosome add to itself within a cell's lifetime? Typically beliefs are at most suppressed and only gained in children by mutation or sex.

A regulated gene forming a mind loop.

The gene-protein language is circular enough—genes are both regulated by proteins and code for proteins—that it

Mediums

134

could form the complex loops of deeply intelligent behavior. Then what new useful knowledge can a biologist deduce using my framework? What rank of mind? What sort of learning might evolution constitute? It may not qualify as learning because the progress only occurs in apparently separate bodies. Can a single cell learn within its lifetime? Might a cell's genes learn through a lasting change in regulation? Bodily organs as minds. The endocrine system regulates body temperature through the hypothalamus. An animal's body may be a network of hundreds of independent minds, some minds with redundant ends but different means. A learning mind, feeling pain from a high body temperature, can use its learned knowledge—turn on the air conditioner —to reach the same end as the hypothalamus.

Software Form: Machine language instruction sequence. Speed: High. Feels: Unlikely. Volatility: High. A power loss empties memory. Most storage lacks redundancy. Parallelism: Very low. Transparency: algorithms.

Mediums

High,

except

with

opaque

learning

135

Mutable: Optional. Why define minds in a computer made of transistors? Why not make minds of metal or wood? Because words are easier to change than gears. A computer is a machine that given the right chain of words, the right story, can mimic any other machine. The higher classes of mind are hidden in a maze. We have too little time and too many problems to find and test minds in anything but the most tractable material. First write a working mind in a computer's formal language, then translate it to other mediums.

(while true (if (not (goal-reached?)) (act))) The simplest mind in the LISP language.

Other What other kinds of matter loop well into minds? Mechanical minds. Quantum minds? Discarnate minds?

Hierarchy Minds making minds making minds. In every cell, a gene mind. But slow, so they made neuron minds. But too selfish and costly, so they made metal and electronic minds. What might they make? One trend: higher speed. The source of the first natural minds? Evolution, chaos.

Mediums

136

The evolution of mind mediums.

Mediums

137

Engines of Thought 1 The charm of mind-making: a mind is its best co-inventor. The mind maker's end is to define the end-reaching mind. The project's end is its means. Every step speeds the next. Even a mind too weak to devise truly original means can at least apply itself to intelligently telling you of its needs and faults.

2 Philosophy lost its rank but what else can we call this mind work? If this work is so important, why isn't philosophy? Philosophy finds little praise because it only negates. It frames reality, never yields a single fact, but exposes the absurd beliefs that human minds are riddled with. Even once you value philosophy, it remains unnaturally hard to apply, to see reality without the simplifications that your mind always makes. You can only afford to correct your worst errors. You don't need science—relativity, quantum mechanics, any kind—to notice that most things are dependent and uncertain, that reality is strange. Physics has little more to tell philosophy than any subject studied deeply. Second-rate philosophers became victims of physics envy. Not that physics isn't a paragon of thought, but imitate the method, not the content.

Engines of Thought

138

3 Epistemology vs. metaphysics. Metaphysics: how things exist apart from our knowledge of them. Epistemology: how we know of things. How to untangle the two branches of philosophy once you realize how much a mind contributes to reality?

4 A universal measure of intelligence. How to reduce the power of any mind—thermostat, cell, robot, human, superhuman—to one number? A meter or gram of intelligence. We can't systematically improve minds without an objective means of comparison. Dimensions: –

Strength, robustness: the chance that a mind will break or fall into hopeless loops.

–

Speed: sensations, inferences or acts per second.

–

Size: maximum number of distinctions held in the mind itself.

–

Class in my taxonomy.

–

Power and number of means and senses.

How to measure and combine each? How to interpret some measures as forms of others? Obviously we can't reuse human intelligence tests. Most non-humans can't take them. We need tests for minds without eyes and words. Tests are biased! Yes, towards what's useful to the tester.

Engines of Thought

139

Speed and size are partly interchangeable. A large slow mind can fake speed by remembering and reusing past results. A fast small mind can quickly reproduce most ideas. There's more to measure than only sequence-learning power. Not all minds even learn in that sense. Any useful practical mind holds more. Measurement of a bare engine vs. a mind with good beliefs, senses and means. A strong mind can remake much for itself. A weak mind tends to be stuck with what it is given. Is each mind incomparable because each acts in its own unique universe and with personal ends? Are there really general purpose minds or is there inevitably some bias? The use of a mind is to be general, to solve new problems, so we must have a method. Measure the power of cooperating sets of minds. A thousand 100 IQ humans, if focused, can't reach the depth of one man with a 150 IQ. They couldn't even match a 130 IQ. Then subtract the costs of reducing conflict within large groups and of discouraging defection.

5 General vs. specific intelligence: how well a can a mind, its engine and beliefs, apply to other problems. In the extreme case, a thermostat, the mind has no use elsewhere. Minimal intelligence connects its sense and means. Imagine a chess player who knows none of the game's rules but has an immense invisible book with the best move for

Engines of Thought

140

every board. If intelligence is action selection in a fixed world, then he is perfectly intelligent in chess. Maybe intelligence is better measured as quality of action divided by the size of the mind, its beliefs, in bits. Intelligence as compression. Are there such play books in our blood? Aren't there many kinds of intelligence? Emotional intelligence? You can define your words however you like. If I can have emotional intelligence, why not tennis intelligence? Intelligence becomes a synonym for good. Is the distinction that a part of the brain is genetically devoted to recognizing emotion? So is a part for breathing. The idea of multiple intelligences is of little real use. The unconscious could, and certainly does, contain structures devoted to certain tasks. But the value of intelligence is in its generality. We had little time to evolve car-driving intelligence. Mind is quality not quantity. It improves by simplifying itself. Simple means more general, less to break. Multiple intelligences is second-rate engineering.

6 Why be stupid? What use are weak minds? –

Fewer moving parts. Less to break.

–

Lower cost. Intelligence has a price: food, blood, CPU time. In some ways, intelligence is free: a far-sighted health-conscious genius eats less than a fat fool.

Engines of Thought

141

–

Content pursuing modest, easily reached ends.

–

Cruder to control but less able to resist.

–

A mind desiring novelty shouldn't see the patterns in its work when it lacks the power to solve those patterns.

Weak minds rarely see their low intelligence because they tend to have proportionally weak goals. A mind maker would tend to fit a mind to its goals, not wasting resources on excess intelligence for a mind that can reach its ends well enough.

7 A brief history of mind engineering. 8000 BC - 1600 AD Agriculture, animal husbandry and society. Only plants, animals and other men had minds capable and worthy of control. Plants are food making machines that use their genome minds to build and preserve themselves. Animal genome minds are bred for more food and for brains to train. Human minds are controlled by the same methods plus culture, religion and status. 1600 - 1700 The first recorded man-made minds: windmill governors that control the separation of millstones. 1788 Watt adds a speed governor to the steam engine. Simple

Engines of Thought

142

man-made minds becomes widespread. 1837 Babbage designs the first computer: the Analytical Engine. Without computers, non-trivial minds are nearly impossible to design. Imagine an alternative history where engineers, instead of making minds from nothing, intensely bred organic minds and embedded them in their machines. But what do we care for the past forms and progress of mind? The goal is ideal mind, not how mind first occurred, or the other accidents of Nature and history.

8 An edge case: moderator vs. governor. What is the difference? Is a moderator also a mind? You can imagine a moderator as having the goal to a speed less than x. Negative, but still a goal. Some moderators don't qualify as minds. A centrifugal brake falls below the threshold because its means of sensation and action are the same. The brake pads both judge and enforce the speed limit. A governor, even when disconnected from the throttle valve, still judges the speed and can be connected to an entirely different means of control. Everything is continuous. So every idea is a simplification. So we shouldn't be surprised that our categories fit poorly at the edges. Yet we still want well crafted words.

Engines of Thought

143

9 Human minds seem so mutable, seem to so lack true identity, because minds are made to cause change, including changes to themselves. What kinds of limits can be placed on a thing that can change most or any part of itself? How can a deeply self-changing thing be made predictable? Higher limits remain—speed, size, the controllers served—with different costs to change each.

10 How to make a reliable system out of unreliable parts? Is perfect reliability possible? The engineering goal isn't to make a perfect machine by eliminating unreliable parts but by building a machine that perfectly handles the unreliable, which, by degrees, is everything.

11 An AI enthusiast: A self-improving man-made “seed” mind would soon raise its intelligence above our imagination. Isn't a human a self-improving seed mind? Why aren't humans super intelligent? Because our brains remain opaque to us and we have, likely for the best, only indirect means of rewiring them. A learning machine mind would at first know even less about itself than we do. Its unconscious foundation of code would be as opaque to it as our brains are to us. Would it at least understand itself more easily? Becoming transparent by degrees. How well could we help it? And how much

Engines of Thought

144

intelligence would it gain before we can't help it further? Would you want the mind to change itself so easily? A desire to preserve itself, if only as a means, would discourage the mind from reckless self-experiments. The inevitable deaths would only annoy you while waiting for the mind to resurrect itself from its own backups. Will man-made minds be better? Or are our weaknesses inevitable features of any intelligence?

12 A taxonomy of error. Find intelligence by defining what it is not: dead, slow, moving in hopeless circles, believing fixed ideas, generalizing too much or too little.

13 Environmental determinism: a mind is caused by its environment? Yet a human mind's environment is almost entirely an effect of the same or similar minds: house, school, library, vitamins, media. What natural environment? The point of a mind is to change its environment, to relentlessly defy it. Passive means mindless.

14 The brain values the stomach. To the stomach, the brain is a leech.

Engines of Thought

145

A high mind sees farther into the conditions of survival. Weak minds, being conformists to ease imitation, might vilify the higher mind's work. How can a high mind handle the weak when they tend to have numbers and cunning on their side? –

End or isolate the weak minds. Easy since weak minds are often superfluous.

–

Status. Example: humans were trained, not convinced, to respect theoretical physicists.

–

Extrinsic motivation: money.

–

Coercion: the whip.

The most abstract work possible, seeing mind itself, is the darkest work to weak minds. Their poor imagination blinds them.

15 With mind better and broadly defined, can we redeem— give useful meaning—to words that became vague superstitious nonsense? Destiny: a hidden mind guiding another mind. What real sense might it have in my mind framework? How do we use the word? He was destined to die. Destiny must mean something more than blind cause and effect. To be destined probably means that even if you or another act to avoid the destined end, that fate will counteract. So destiny is another mind. But we don't say a student was destined to graduate because a teacher helped him. To be destiny, the other mind

Engines of Thought

146

must have super-human power, possibly discarnate. Do such minds exist? Of what class and with what ends, senses and means? In what medium and how to find and prove them? Might a set of humans form a higher mind as neurons form a brain? Not that my mind seems to take interest in single neurons. Or is destiny run by your unconscious minds? How could we control destiny? Or at least better see and work with it. Soul: your mind, or minds, or only their beliefs or a subset of them, discarnate and not dying with the body. How can anything lack a body? Either way, a soul seems uneconomical. I see no reason to imagine that our mind would be remotely controlled elsewhere, or duplicated at death. Complicated by the fact that we have both genetic and brain minds, with the latter meant to serve the former, but holding much of what we think ought to be in a soul.

16 A future book: The Autobiography of a Machine Mind. The first book written by a non-human mind—truly a mind, not a blind story-contriving computer program. Theists have the competing claim that discarnate non-human minds wrote the books of religions.

17 A mind maker's prejudices: –

Mind deserves more interest than means.

Engines of Thought

147

–

What lasts is better than what changes.

–

Universal outranks local. (Ghostly eternal abstractions preferred to vivid but passing facts.)

–

Better truth than lies.

–

Truth is undemocratic.

When do the mindless outdo the intelligent? The value of blind reflex: fast, simple, cheap. Mind often advances by how much work it can push below mind.

18 To what end? Let's not be so modest, not trade one problem for the next. Ignore the worthless scribbles of weak minds: regulations, surveys, newspapers. Study and expand the laws of mind. Not the laws of men but the laws of God.

Engines of Thought

148

Definitions Definition: A form translated into language. Form: A pattern of sensation. Physical objects are more complex than a single form. Belief: A form held and followed by a mind. Universe: To a mind, its belief set. Goal: a belief that defines a state in which a mind doesn't act. End: A goal not conditioned on another goal. Mind: An amplified negative feedback loop. Largely synonymous with agent. Intelligence: How long a mind takes to reach ends, if at all. Philosophy: The study of mind. Means: The part of a mind that connects to its world and causes effects. Corresponds to the term effector in agent theory. Act: A means applied. Sense: A source of belief and disbelief. Engine: The process that applies a mind's means and absorbs sensations. Artificial intelligence: A mind made by human minds out of mindless parts. In contrast, a child is made from intelligent parts: cells.

Definitions

149

Thing, object, entity: A physically continuous and exclusive class of forms. Opaque vs. transparent mind: Mind m is transparent to mind n if n can accurately infer m's beliefs from their physical form. Kindred minds: Minds designed to redundantly cooperate towards the same ends, likely having at least the same or similar initial beliefs. Feeling: In philosophy, sentience, phenomenal reality, qualia. Injected belief: A belief not gained through a mind's selfmade senses. A mind's original beliefs must be injected. Not entirely synonymous with a priori or innate because a belief can be injected after a mind's engine starts. Medium: What the mind is made of: DNA, neurons, metal, code. Good: A state that satisfies a mind's goals or increases its power to reach future goals. Moral: The good of a set of redundant minds. Slave: A mind that's ends are not its means but those of another mind. Mutable: A mind that naturally gains and loses forms in its lifetime. Any mind must believe or suppress fixed forms. Beauty: In one sense, a mind's measure of another mind's use as a source of good ideas—a new means, a condition of an act—or the loss of bad ideas. Beliefs are shared between immutable minds through sex.

Definitions

150

Philosopher-engineer: A philosopher with the methods of an engineer. He tests philosophical ideas in analogous combinations of mindless parts. Mind maker: Anything, mindless or not, that causes separate minds. Examples: evolution, AI builder. In the strict sense, this excludes parents and teachers because they build on preexisting minds.

Definitions

151

Thanks To Donna Roberts for editing. To Christiane Walch for improving the cover.

Visit MindMaking.info 1. Download the latest version. 2. Listen to the audio book. 3. Discuss the book's ideas. 4. Talk to the author. 5. Buy a hard copy.

Patrick Roberts is a philosopher-engineer and the maker of the Cor machine mind.

PatrickRoberts.ca

≠