Showing posts with label illusion. Show all posts
Showing posts with label illusion. Show all posts

Monday, April 3, 2017

Understanding "Talking Lions" and Other "Black Boxes"


If a lion could speak, we could not understand him.
-- Wittgenstein[1]


Artificial intelligence is everywhere. But before scientists trust it, they first need to understand how machines learn.
-- D Castelvecchi, Can We Open the Black Box of AI? [2]


"With artificial intelligence, we are summoning the demon."
-- Elon Musk (2014) Reported in CNN Tech , Oct 28th.


 
Faust with Homunculus

Black Boxes Abound. Less so, our understandings of them. The "black boxes" are not only lions, or AI devices, but also humans, and even more. If we generalize, keeping in consideration our perceptual limitations, any source of emissions that appear to us to be more than random, from any object or even an "empty" focus of attention, is, initially, a black box.

Many commentators have remarked that Wittgenstein makes an incautious jump from speaking to understanding in the quote cited above. I will generalize their concerns: how do we know that a (what appears to us to be) non-random emission of, say, sound, is speech? Or, if the transmission is not sonic to humans, language? Animals can learn to mimic human emissions. Even non-English recordings, played backwards, can sound like English speech.

To many a monolingual English speaker, the following phrases sound like pieces of English nursery rhymes colored with a non-English accent:
1. French - Un petit d'un petit s'étonnent aux Halles (Humpty Dumpty sat on a wall.)
2. German - Oh wer, oh wer ist Mai Lido doch Gong? (Oh where, oh where is my little dog gone?)
3. Spanish (Caribbean) -- Grima! Sí, comí. Te excusé que rifa. (Christmas is coming, the goose is getting fat.) [3]
Black Boxes are "Understandings." How does Wittgenstein even judge what the lion is doing? What assumptions is he making? Even if the lion's emissions sound like an English sentence, what is relevant in the context? What is the lion doing? Suppose, for example, our friend, Harry, is standing before a lion's cage in which a cave has been constructed as a den. The lion, staring at Harry, emits the sound sequence, /2aym+ gowing+3 hówm1/ [4]

It sounds like the lion just said, "I'm going home." But did it "say" that? Suppose the lion then lies down and rolls over on its back. It starts to snore. Are we still inclined to think that the lion talked? (If so, did he talk to Harry? Did the lion inform Harry it was "going home?" And, thus, was the lion telling a lie?) Can Harry believe his eyes and ears? Has he jumped to conclusions? (See Artificial Intelligence Weirdness. Need categorizing relate to visual cues?)

The basic problem is how we distinguish illusions, visual, as well as audial or other, from realities. Even AI's have this problem.(See Autonomous Car Collides with Bus: an illusion of abstractions?)

And What Are Understandings? Understandings, when articulated in language, are narratives (or, if highly structured, a program) of collection or connection. We may judge them incorrect, false, or incomplete but they are still understandings -- we might say, "misunderstandings."

People may not be at all articulate as to what they understand so we may have to observe how they proceed from their present conditions toward those outcomes that we judge them to be pursuing. If we observe persistent failure to achieve a goal we might reasonably judge that their understanding of how to achieve that goal is deficient.

There is a large number of likely misunderstandings that persist in any population because they need not or cannot often be put to any test whose outcomes enjoy broad consensus as to their pertinence. (So what if kids believe in the Tooth Fairy!) These questionable understandings are often characterized as (empirically) "non-disconfirmable" beliefs. Examples are:
a. The mind is co-extensive with the body, or:
b. The mind is not coextensive with the body.
c. The universe is a hologram; or
d. The universe is not a hologram.

It would appear that any rationales, or chain of rationales, that contain misunderstandings would thereby be severely weakened. (Many pundits who presume to speak for Science share such misunderstandings with those who presume to speak for Religion. See Pseudo-Science: the reasonable constraints of Empiricism.)

The Fractal(?) Generation of Rationales. Understanding in general is the ability to produce chains of behavior or their narratives which link a confronted situation of interest to a goal to be achieved. But, understandings alone are often narrow and give may give no indication of the interests or abilities of the persons who understand and may be expected to act. Thus, understandings may be just part of what we're looking for.

Rationales can be elaborated from understandings. Rationales often bring up systemic concerns about, say, indicator validity (Cue), actor interests (Concern), and actor abilities (Control) which underly attempts at interventions based on narrower technical understandings.

Understandings can be chained together to produce rationales for action, and rationales themselves chained to produce broader understandings. Thus we may go from merely speculative understanding of how to cross a river, to a rationale for expanding commerce across that river, to an understanding of how political influence can be brought about reiteratively by using market rather than, say, armed forces. Understandings developed this way can be connected together, in chains or trees, etc., to articulate, say, foreign economic development policies. (See The Fractalization of Social Enterprise)


Why Can We Understand Human Black Boxes but find It harder to Understand AI Black Boxes?
It is because many people working in AI feel restricted to technical issues of understanding the how's of economically pertinant AI functioning. Despite the the persistent anthropomorphizing of AI output -- and despite the suspicions that the ultimate goal of AI research is to create some kind of homunculus -- issues of rationale, especially in the contexts of human development and learning are discounted as off target. [5]

Autonomous AI stimulates the same kind of misgivings that one might feel about extra-terrestrials. We are not sure we can predict, much less control what they might do with us. It is a non-sequitur to believe that high computational ability equates to altruism. (See METI. Here We Are! Come Eat Us! Our Children Are Especially Tasty!)

We have built-in, so to speak, Cue-identifying abilities inherited through evolution, both physical, mental and social, that are sensitive to the norms of the environments of our development. We are Concerned to share a world with motile, sometimes dangerous, beings that we can hear, feel, smell, taste as well as see. And we know from personal experience what fear, hunger, danger, hate and social attraction are. [6]

We measure the autonomy we are willing to extend to our ancient non-human friends and enemies because we can guess well what drives them, and how to accommodate them to our society. And, not the least important, we can, with more or less success, Control and defend ourselves against them.

To pursue the issues of understanding and rationale, see Intervention. Helping, interfering or just being useless?

Cordially, EGR

(P.S. Check out this interesting interview with Gary Marcus called Making AI More Human.)

NOTES
[1] L Wittgenstein 1958b. Philosophical Investigations. ed. G.E.M. Anscombe and R.Rhees. tr. G.E.M. Anscombe, 2nd edition. Oxford: Blackwell. page 223. See comments by Simon van Rysewyk at Wittgenstein Light

[2] D Castelvecchi Can We Open the Black Box of AI? Nature magazine, October 5, 2016

[3] Sources for pseudo-English concoctions: French - Mots d'Heures: Gousses, Rames, The d'Antin Manuscript. Penguin,1980; German - J Hulme Mörder, Guss Reims. The Gustave Leberwurst Manuscript, Clarkson N Potter, 1981; Spanish -- My confabulation, EGR

[4] IPA, digits indicate tone levels. Cf. H A Gleason, An Introduction to Descriptive Linguistics (New York: Holt, Rinehart and Winston, 1955)

[5] See EG Rozycki Behavior in Measurability and Educational Concerns 

[6] See D Gross, Why Artificial Intelligence Needs Some Emotional Intelligence

Friday, May 27, 2016

Autonomous Car Collides with Bus: an illusion of abstractions?

The secret of success is honesty and fair dealing. If you can fake those, you've got it made. -- Groucho Marx
re-edited 3/12/16

An Illusion is not Necessarily a Perceptual Error. It is usually just a rather run-of-the-mill experience. We encounter illusions frequently. To our personal benefit, our own bodies often produce illusions, e.g. 3D vision or directional hearing, or even the pains we experience from surprising external or internal systemic pressures, or, even, upon hearing bad news. The pain in our bare foot is not the tack we just stepped on; nor, the pleasure we feel, the chocolate we just ate.

We humans rely on illusions. We use them for better or worse to lead our lives. Seafarers have long guided themselves across oceans using star constellations. But we know now that there’s no Big Dipper, or Belt of Orion, to be found “out there.” Constellations of stars are illusions due to slow cosmic change and to our position as viewers. Human racial differences, too, have been a commonly encountered practically exploitable illusion due to slow change and social position. (See Philosophy & Racial Discourse.)
Wayfarers and pilots, too, may venture into or over deserts and forests, cross-referencing their positions along the way by triangulating imaginary lines-of-sight between “fixed” terrain features. (Even such “fixtures” as expected mirages of an oasis or other bodies of water can work.)

We humans are not unique in producing illusions in pursuit of the “goals” patterned in our genes. Our use of duck calls, fishing lures or pheromone hunting baits illustrates this. Animal species, too, may pursue prey or fend off predators by, feigning death, or, ferocity, or venomousness.

Illusion is not in the perception, in the appearance, but in what we take that appearance to be. We learn at a young age “Appearances can be deceiving.” We certainly hope so, when we aggressively yell to frighten off an approaching animal; or, when we go to meet someone on a first date.

How do we learn what is illusory and what not? By using as many of our senses as we can, in as many variations of position as we can manage, in order to see what happens. It is no accident that most illusionists do their presentations to audiences fixed in their seats. Or that séances are held in the dark. Or that proselytizers of every stripe discourage broad learning in their followers. Narrow perspectives blur the distinction between illusions and realities.

Not only do we vary the manner in which we confront what we suspect may be illusory, but we do it noting the context of presentation. Is it “normal”? Is the lighting different from daylight? Would art museums display their traditional exhibits under intense red light? (Try sorting blue, black and brown socks under incandescent light. Why is fluorescent light better for the task?)

Is image, tone, feel, odor, or taste sufficiently persistent? How far away (intense) does the apparition, tone, etc. seem to be? Smallness and haziness are indicators of distance. If we can estimate its height (or loudness, etc.), then, more is closer.

Strained or strange body positions on the part of the observer are likely to indicate something amiss. Cross a middle finger under a forefinger, close you eyes and touch the tip of you nose with both fingers simultaneously. You will feel two noses!

We tend to judge the reality of the perception that confronts us based on the degree of “normal” corroboration among our different sense modalities in the context of our estimation of the “normality” of the perceptual situation.

Is the image, tone, feel, odor, or taste sufficiently persistent? Can we compare it with tracings from memory, or with our hypothetical constructions about confrontations to come? Both past and future are processings left as mental traces, not the evanescent qualia that phenomenologists say present themselves to us only in the passing moment. (See Similarity: not a given; but, a composition.).

Motion pictures, or their derivatives, provide us an interesting example of illusion. They are composed of stills, or frames, the projections of which are clearly illusions, which appear less illusory when perceived quickly and sequentially. Their most important consequence is the illusion of cause and effect.

Cause & Effect. What conditions “connect” perceptual events into perceived chains of cause and effect? This feature seems to be “wired into” us; (or, as Immanuel Kant proposed much earlier, cause-effect is a essential structural component of Mind - Verstand) Cause-Effect is often an illusion we must take pains to recognize as such; e.g. Post Hoc, Propter Hoc. (See Cause = Perceptions + Assumptions )

How do we correct a suspected misconception as the reality of a perception? By probing any hunch – called in exalted educational circles, “testing our hypothesis” – to see if it holds up under questioning. Usually fatigue, boredom and feelings of risk set in long before any “personal hypothesis” is thoroughly criticized. (See Good Lies, Wise Evasions )

Reality, Knowledge? Just Presumptions? Some long while ago, at a cocktail party, in my late youth and early marriage, a chemistry professor, having had a bit too much chemical and having heard from my in-laws that I was still in my middle-twenties (!) stumbling around after a profession, confided in me: “Consider Chemistry! By the year 2000 we will be all efficiently taking various pills instead of eating, having to waste our time and effort chewing, so-called-healthy crunchies for hors d’oeuvres.”

Granted, time passes, new things are discovered, the old dies away. Today’s knowledge becomes tomorrow’s superstition; today’s realities, tomorrow’s illusions. This seems inevitable in a culture which practices an empirically-based knowledge acquisition model.

An Equation-like Metaphor. But this is no argument for returning to a culture of unquestionable, alledgedly prophesy-based dogmatism. Rather, think of “reality, R,” as a limit condition, a congeries, at least, of truths, T. The disconfirmation of hypotheses, as a procedure (a function), D, which moves us away from previous illusions and toward “reality, ” as the number of disconfirmations, D, increases, likely to infinity (since we and the universe around us evolve). Knowledge (i.e. Compiled Truths, {T}, which will in the limit comprise Reality, is the residue of practical, experimentally exercised caution. See Knowledge: The Residues of Practical Caution..

A metaphorical equation: (lim K/(D → ∞)) = {T} ⊆ {R}

Read this: In the limit, our knowledge, K, as the disconfirmation of tentative hypotheses increases to infinity, approaches - via {T} - Reality (R),


a. T is the set of all truths, {T}, i.e. hypotheses that have survived disconfirmation tests;


b. {T} is reasonably expected by less-than-omniscient beings such as us, to be always less than (R).
I concede this to be a metaphor because I am not even sure it makes much sense taken literally. (See Measuring Educational Outcomes: truth, tricks and hype.) Certainly the total number of English (or other language) words (and the power sets of all their combinations and permutations) is less than the number of events discriminable by us in the universe as we know it, or even as we might come to know it in the future, given superior instrumentation and language to expand and communicate our experience.

Contingently Dysfunctional Characteristics: explaining AI failure. I do not have enough particulars to deal with the specific recent collision event. But I do have a broader concern that may apply to it. It is that there is a problem with the traditional notion of abstraction used by (only some, I hope) AI researchers.

The Platonic conception of abstraction introduced by Socrates in the Meno – the traditional use of necessary and sufficient (N&S) conditions -- common to many theoretic activities to this day, appears in full bloom in chapter two, “A Model of the Neo-Cortex,” of Ray Kurzweil’s book, How to Create a Mind.

(Compare Alfred Korzybski's "Structural Differential," shown above - from Science & Sanity: An Introduction to Non-Aristotelian Systems and General Semantics (1941) - that he used to explain his conception of abstraction.)

This model of abstraction, call it the N&S model, is, what I will call a “ context pro-functional” accumulator. It sets up a checkpoint for the selection of candidates in a given class of situation, call it [S1], to moving up the ladder of abstraction. It allows items to climb that meet certain functional requirements for an AI construction goal. It may even cull out an [S1]-dys-functional elements, if they are of concern.

But such dysfunctional elements may be overlooked or disregarded, as being, for example of low probability of happening, or of low enough cost in the long run to allow to pass through. See Concept as Abstraction. A hindrance in developing intelligence?

But the [S1]-appropriate abstraction may also be a sort of “sludge-accumulator,” in that it allows to pass through any items with [Sn?1]-dysfunctional characteristics so long as they are [S1]-functional. A situation [S31] may arise in which the response set available contains [S31]-dysfunctional items that are nonetheless [S1]-functional.

So it is that an autonomous car might accelerate to pass a bus that has lain quiescent for twelve seconds, [S31], but may accelerate on eight because the bus driver left his running lights off, creating in the car an "illusion" of [S1], signaling a permanent stop.

Using the common N&S concept of abstraction is very much like an admissions procedure to an organization based on education, ethnicity, profession and income that welcomes anyone who meets these four requirements but ignores, say, their legal status, e.g. felon or not, their level of indebtedness, their history of violent public behavior, and the condition of their health.

How to Avoid Such Difficulties Basically it comes down to three things:

a. investigating broad ranges of functional situations;

b. looking at not only those characteristics that are pro-functional, but considering whether there are potentially dysfunctional characteristics that may manifest themselves in not impossibly remote circumstances;

c. building in different kinds of input systems to enable corroboration among them.
Scanners alone, like visitors' eyes at a seance, will not easily help their user recognize illusions.

For examples and to pursue these issues, see Physical Objects as Non-Individuated Entities. (A universe of constructs) )

Cordially
--- EGR


Thursday, December 19, 2013

Cause & Effect: essential belief, misapprehension, moral hazard?

updated 071322
The brain is a construction of the brain. -- Nietzsche
There Must Be a Better Way! (The proctological principle?) I was waiting in a doctor’s office to arrange an examination, when a nurse answered a phone call. The irate voice on the other end of the line resounded throughout the small room: “I’m getting a colonoscopic exam tomorrow and I’m told I have to flush my system out by drinking a gallon of some insipid water concoction over the next twenty-four hours.” “That’s correct,” replied the nurse. “That’s absolutely ridiculous!” retorted the caller. “In this day and age there must be a better way to do it!” “No, sir. There is not,” responded the nurse.

A stream of expletives gushed from the telephone receiver. I thought to suggest to the nurse that she advise the patient to use a plumber’s snake instead of the prescribed water solution. But I kept my silence.

Practically Necessary Beliefs. Our way of life depends on a certain system of beliefs handed down through the centuries, although substantially modified in the process. Our notions of self-control, insanity, responsibility, criminality, and justice rest ultimately on the belief that some events relate to one another as cause and effect.

Daily we debate about who has done what; or, who can do what -- pretending all along that these “practical” questions have little to do with what “ivory-tower types” would call “philosophy.” In the debate we presume our investigations rest on a firm foundation: not merely the “belief,” but the “reality” of cause and effect. The operative train of thought here seems to ride on the Proctological Principle: if we (think we) need it, it must, in this day and age, be “a reality.”

Are Cause and Effect Illusions? Bertrand Russell recounts in his memoirs how, as a young man, he woke up in his bed one morning and came to a frightening understanding as to what David Hume meant by his theory of causation. All we ever observe, Hume had written, are correlations; regular sequences of event. Cause is our interpretation, our belief, more commonly, an assumption that normally observed regularities of precedence and propinquity will continue into the future.

Russell feared to step out of his bed onto the floor. It might, if Hume were right, suddenly open up beneath him, since there was nothing more than assumption to guarantee the persistence of its support. Thus, he would fall into an abyss of … nothingness (?).

(Russell’s reasoning here is somewhat inconsistent: If he could not trust that the floor, though sound and normal, would continue to support him, how could he come to any conclusion about what might happen next? Maybe he might float over the abyss!)

Being ultimately a pragmatist, as we all tend to be, Russell got past this fear and out of his bed and on with his life. Possibly he came to the realization that his bed, too, -- or the foundation of his house, or England in its entirety -- could just as likely open up to swallow him as he imagined the floor might.

Ultimately, perhaps, Russell, being a philosopher, might have come to concede (he did so, later) that there is more involved in identifying cause-and-effect relations than what our perceptions give us. Even today we warn our students, friends, associates, and acquaintances, perhaps, -- often futilely -- that mere “correlation” neither indicates nor denies “cause.” Little matter; if they are gamblers, day-traders, newspaper columnists, or political incumbents, they tend to ignore us.

Cause = Perceptions + Assumptions. Russell’s, and our, ability to act in this world rests heavily on structures of trust in generally unproven (and purely empirically unprovable) regularities of persistence. We construct edifices of trust with bricks of hope and desire cemented together with no more than the mortar of our assumptions. (See Varieties of Explanation & Cause.)

Immanuel Kant, some decades after Hume, presented a counter to Hume’s apparent empirical skepticism by denying that the human mind could go beyond experience, even such experience as augmented with scientific apparatus, to settle questions such as the “nature” of causation or objectivity. Our knowledge, Kant says, such as it is, is limited to phenomena.

Phenomena are constructed within the framework of human categories of understanding, which constitute the structure of the human mind, although interacting in someway with entities -- or entity -- of the extra-phenomenal world. Number, unity or plenty, says Kant, is a category of human understanding; not necessarily to be expected to parallel extra-phenomenal “reality.” (Check out the notion of quantum entanglement to appreciate how Kant's distinctions still retain relevance; although, he seems not to have considered how developing technologies and theories might interact to substantially expand human experience.)


It is amazing how many people in so many walks of life just hate this kind of theory: the theory that in culling out the true from the false, the wishful or delusional from the real, we ultimately invoke items of belief that themselves can only be judged, ad infinitum, by invoking other items of belief. (See Getting Down to Facts: can we avoid making assumptions?)

Is there a Bedrock of Knowledge? So what if claims of knowledge are always risky, resting on a potentially infinite chain of assumptions? Why the panic? Or more likely, the dismissive disdain? (Much more polite, if more condescending.)

I suppose it’s frightening. What? No absolute bedrock of The Known upon which to build our temples of Value, Prestige, Justice and Faith? Must everything be at risk? Can we, no matter how high our IQ or repute, how vast our wealth or influence, never be free from the risk of mistake?

Excepting formal linguistically constructed systems, i.e. knowledge by definition or stipulation, never, it would seem.

But does it really matter whether we occasionally stumble through life, rather than exhibiting perfection in all our decisions? Apparently to some people it does. Many who would reject the idea that knowledge may be essentially tentative are not inclined to even bother to present counter-argument or -evidence. Their basic concern, one might suspect, is more to preserve their incumbency in privileged positions; or, their right to importune others less empowered, i.e. to be in a condition of permanent moral hazard.

For references and to examine these issues further, See Buffering: Enhancing Moral Hazard in Decision-Making?

See also Varieties of Explanation. & Cause

Cordially
--- EGR

**Consider the similarities also in
Isaiah 45:7 (KJV) "I form the light, and create darkness: I make peace, and create evil: I the Lord do all these things."

And, also, the conversation between Faust and Mephistopheles (1335-1337 and 1345-1358 Original German edition, my translation -- EGR) wherein Faust's question as to who Mephisto is is answered "I am a part of that power that always pursues Evil yet always yields Good."

Faust counters, "How can you be a part of something when you present yourself to me as a whole?", Mephisto replies,"... Although (you) a human, from that a tiny world of fools, may consider yourself a whole, I am a part of a part that once was Everything, the Darkness that gave birth to Light."