Archive
Joscha Bach: We need to understand the nature of AI to understand who we are
This is a great, two hour interview between Joscha Bach and Nikola Danaylov (aka Socrates): https://www.singularityweblog.com/joscha-bach/
Below is a partial (and paraphrased) transcription of the first hour, interspersed with my comments. I intend to do the second hour soon.
00:00 – 10:00 Personal background & Introduction
Please watch or listen to the podcast as there is a lot that is omitted here. I’m focusing on only the parts of the conversation which are directly related to what I want to talk about.
6:08 Joscha Bach – Our null hypothesis from Western philosophy still seems to be supernatural beings, dualism, etc. This is why many reject AI as ridiculous and unlikely – not because they don’t see that we are biological computers and that the universe is probably mechanical (mechanical theory gives good predictions), but because deep down we still have the null hypothesis that the universe is somehow supernatural and we are the most supernatural things in it. Science has been pushing back, but in this area we have not accepted it yet.
6:56 Nikola Danaylov – Are we machines/algorithms?
JB – Organisms have algorithms and are definitely machines. An algorithm is a set of rules that can be probabilistic or deterministic, and make it possible to change representational states in order to compute a function. A machine is a system that can change states in non-random ways, and also revisit earlier states (stay in a particular state space, potentially making it a system). A system can be described by drawing a fence around its state space.
CW – We should keep in mind that computer science itself begins with a set of assumptions which are abstract and rational (representational ‘states’, ‘compute’, ‘function’) rather than concrete and empirical. What is required for a ‘state’ to exist? What is the minimum essential property that could allow states to be ‘represented’ as other states? How does presentation work in the first place? Can either presentation or representation exist without some super-physical capacity for sense and sense-making? I don’t think that it can.
This becomes important as we scale up from the elemental level to AI since if we have already assumed that an electrical charge or mechanical motion carries a capacity for sense and sense-making, we are committing the fallacy of begging the question if carry that assumption over to complex mechanical systems. If we don’t assume any sensing or sense-making on the elemental level, then we have the hard problem of consciousness…an explanatory gap between complex objects moving blindly in public space to aesthetically and semantically rendered phenomenal experiences.
I think that if we are going to meaningfully refer to ‘states’ as physical, then we should err on the conservative side and think only in terms of those uncontroversially physical properties such as location, size, shape, and motion. Even concepts such as charge, mass, force, and field can be reduced to variations in the way that objects or particles move.
Representation, however, is semiotic. It requires some kind of abstract conceptual link between two states (abstract/intangible or concrete/tangible) which is consciously used as a ‘sign’ or ‘signal’ to re-present the other. This conceptual link cannot be concrete or tangible. Physical structures can be linked to one another, but that link has to be physical, not representational. For one physical shape or substance to influence another they have to be causally engaged by proximity or entanglement. If we assume that a structure is able to carry semantic information such as ‘models’ or purposes, we can’t call that structure ‘physical’ without making an unscientific assumption. In a purely physical or mechanical world, any representation would be redundant and implausible by Occam’s Razor. A self-driving car wouldn’t need a dashboard. I call this the “Hard Problem of Signaling”. There is an explanatory gap between probabilistic/deterministic state changes and the application of any semantic significance to them or their relation. Semantics are only usable if a system can be overridden by something like awareness and intention. Without that, there need not be any decoding of physical events into signs or meanings, the physical events themselves are doing all that is required.
10:00 – 20:00
JB – [Talking about art and life], “The arts are the cuckoo child of life.” Life is about evolution, which is about eating and getting eaten by monsters. If evolution reaches its global optimum, it will be the perfect devourer. Able to digest anything and turn it into a structure to perpetuate itself, as long as the local puddle of negentropy is available. Fascism is a mode of organization of society where the individual is a cell in a super-organism, and the value of the individual is exactly its contribution to the super-organism. When the contribution is negative, then the super-organism kills it. It’s a competition against other super-organisms that is totally brutal. [He doesn’t like Fascism because it’s going to kill a lot of minds he likes :)].
12:46 – 14:12 JB – The arts are slightly different. They are a mutation that is arguably not completely adaptive. People fall in love with their mental representation/modeling function and try to capture their conscious state for its own sake. An artist eats to make art. A normal person makes art to eat. Scientists can be like artists also in that way. For a brief moment in the universe there are planetary surfaces and negentropy gradients that allow for the creation of structure and some brief flashes of consciousness in the vast darkness. In these brief flashes of consciousness it can reflect the universe and maybe even figure out what it is. It’s the only chance that we have.
CW – If nature were purely mechanical, and conscious states are purely statistical hierarchies, why would any such process fall in love with itself?
JB – [Mentions global warming and how we may have been locked into this doomed trajectory since the industrial revolution. Talks about the problems of academic philosophy where practical concerns of having a career constrict the opportunities to contribute to philosophy except in a nearly insignificant way].
KD – How do you define philosophy?
CW – I thought of nature this way for many years, but I eventually became curious about a different hypothesis. Suppose we invert our the foreground/background relationship of conscious experience and existence that we assume. While silicon atoms and galaxies don’t seem conscious to us, the way that our consciousness renders them may reflect more their unfamiliarity and distance from our own scale of perception. Even just speeding up or slowing down these material structures would make their status as unconscious or non-living a bit more questionable. If a person’s body grew in a geological timescale rather than a zoological timescale, we might have a hard time seeing them as alive or conscious.
Rather than presuming a uniform, universal timescale for all events, it is possible that time is a quality which does not exist only as an experienced relation between experiences, and which contracts and dilates relative to the quality of that experience and the relation between all experiences. We get a hint of this possibility when we notice that time seems to crawl or fly by in relation to our level of enjoyment of that time. Five seconds of hard exercise can seem like several minutes of normal-baseline experience, while two hours in good conversation can seem to slip away in a matter of 30 baseline minutes. Dreams give us another glimpse into timescale relativity, as some dreams can be experienced as going on for an arbitrarily long time, complete with long term memories that appear to have been spontaneously confabulated upon waking.
When we assume a uniform universal timescale, we may be cheating ourselves out of our own significance. It’s like a political map of the United States, where geographically it appears that almost the entire country votes ‘red’. We have to distort the geography of the map to honor the significance of population density, and when we do, the picture is much more balanced.
The universe of course is unimaginably vast and ancient *in our frame and rate of perception* but that does not mean that this sense of vastness of scale and duration would be conserved in the absence of frames of perception that are much smaller and briefer by comparison. It may be that the entire first five billion (human) years were a perceived event that is comparable to one of our years in its own (native) frame. There were no tiny creatures living on the surfaces of planets to define the stars as moving slowly, so that period of time, if it was rendered aesthetically at all, may have been rendered as something more like music or emotions than visible objects in space.
Carrying this over to the art vs evolution context, when we adjust the geographic map of cosmological time, the entire universe becomes an experience with varying degrees and qualities of awareness. Rather than vast eons of boring patterns, there would be more of a balance between novelty and repetition. It may be that the grand thesis of the universe is art instead of mechanism, but it may use a modulation between the thesis (art) and antithesis (mechanism) to achieve a phenomenon which is perpetually hungry for itself. The fascist dinosaurs don’t always win. Sometimes the furry mammals inherit the Earth. I don’t think we can rule out the idea that nature is art, even though it is a challenging masterpiece of art which masks and inverts its artistic nature for contrasting effects. It may be the case that our lifespans put our experience closer to the mechanistic grain of the canvas and that seeing the significance of the totality would require a much longer window of perception.
There are empirical hints within our own experience which can help us understand why consciousness rather than mechanism is the absolute thesis. For example, while brightness and darkness are superficially seen as opposites, they are both visible sights. There is no darkness but an interruption of sight/brightness. There is no silence but a period of hearing between sounds. No nothingness but a localized absence of somethings. In this model of nature, there would be a background super-thesis which is not a pre-big-bang nothingness, but rather closer to the opposite; a boundaryless totality of experience which fractures and reunites itself in ever more complex ways. Like the growth of a brain from a single cell, the universal experience seems to generate more using themes of dialectic modulation of aesthetic qualities.
Astrophysics appears as the first antithesis to the super-thesis – a radically diminished palette of mathematical geometries and deterministic/probabilistic transactions.
Geochemistry recapitulates and opposes astrophysics, with its palette of solids, liquids, gas, metallic conductors and glass-like insulators, animating geometry into fluid-dynamic condensations and sedimented worlds.
The next layer, Biogenetic realm precipitates as of synthesis between the dialectic of properties given by solids, liquids, and gas; hydrocarbons and amino polypeptides.
Cells appear as a kind of recapitulation of the big bang – something that is not just a story about the universe, but about a micro-universe struggling in opposition to a surrounding universe.
Multi-cellular organisms sort of turn the cell topology inside out, and then vertebrates recapitulate one kind of marine organism within a bony, muscular, hair-skinned terrestrial organism.
The human experience recapitulates all of the previous/concurrent levels, as both a zoological>biological>organic>geochemical>astrophysical structure and the subjective antithesis…a fugue of intangible feelings, thoughts, sensations, memories, ideas, hopes, dreams, etc that run orthogonal to the life of the body, as a direct participant as well as a detached observer. There are many metaphors from mystical traditions that hint at this self-similar, dialectic diffraction. The mandala, the labyrinth, the Kabbalistic concept of tzimtzum, the Taijitu symbol, Net of Indra etc. The use of stained glass in the great European cathedral windows is particularly rich symbolically, as it uses the physical matter of the window as explicitly negative filter – subtracting from or masking the unity of sunlight.
This is in direct opposition to the mechanistic view of brain as collection of cells that somehow generate hallucinatory models or simulations of unexperienced physical states. There are serious problems with this view. The binding problem, the hard problem, Loschmidt’s paradox (the problem of initial negentropy in a thermodynamically closed universe of increasing entropy), to name three. In the diffractive-experiential view that I suggest, it is emptiness and isolation which are like the leaded boundaries between the colored panes of glass of the Rose Window. Appearances of entropy and nothingness become the locally useful antithesis to the super-thesis holos, which is the absolute fullness of experience and novelty. Our human subjectivity is only one complex example of how experience is braided and looped within itself…a kind of turducken of dialectically diffracted experiential labyrinths nested within each other – not just spatially and temporally, but qualitatively and aesthetically.
If I am modeling Joscha’s view correctly, he might say that this model is simply a kind of psychological test pattern – a way that the simulation that we experience as ourselves exposes its early architecture to itself. He might say this is a feature/bug of my Russian-Jewish mind ;). To that, I say perhaps, but there are some hints that it may be more universal:
Special Relativity
Quantum Mechanics
Gödel’s Incompleteness
These have revolutionized our picture of the world precisely because they point to a fundamental nature of matter and math as plastic and participatory…transformative as well as formal. Add to that the appearance of novelty…idiopathic presentations of color and pattern, human personhood, historical zeitgeists, food, music, etc. The universe is not merely regurgitating its own noise in ever more tedious ways, it is constantly reinventing reinvention. As nothingness can only be a gap between somethings, so too can generic, repeating pattern variations only be a multiplication of utterly novel and unique patterns. The universe must be creative and utterly improbable before it can become deterministic and probabilistic. It must be something that creates rules before it can follow them.
Joscha’s existential pessimism may be true locally, but that may be a necessary appearance; a kind of gravitational fee that all experiences have to pay to support the magnificence of the totality.
20:00 – 30:00
JB – Philosophy is, in a way, the search for the global optimum of the modeling function. Epistemology – what can be known, what is truth; Ontology – what is the stuff that exists, Metaphysics – the systems that we have to describe things; Ethics – What should we do? The first rule of rational epistemology was discovered by Francis Bacon in 1620 “The strengths of your confidence in your belief must equal the weight of the evidence in support of it.”. You must apply that recursively, until you resolve the priors of every belief and your belief system becomes self contained. To believe stops being a verb. There is no more relationships to identifications that you arbitrarily set. It’s a mathematical, axiomatic system. Mathematics is the basis of all languages, not just the natural languages.
CW – Re: Language, what about imitation and gesture? They don’t seem meaningfully mathematical.
Hilbert stumbled on problems with infinities, with set theory revealing infinite sets that contains themselves and all of its subsets, so that they don’t have the same number of members as themselves. He asked mathematicians to build an interpreter or computer made from any mathematics that can run all of mathematics. Godel and Turing showed this was not possible, and that the computer would crash. Mathematics is still reeling from this shock. They figured out that all universal computers have the same power. They use a set of rules that contains itself and can compute anything that can be computed, as well as any/all universal computers.
They then figured out that our minds are probably in the class of universal computers, not in the class of mathematical systems. Penrose doesn’t know [or agree with?] this and thinks that our minds are mathematical but can do things that computers cannot do. The big hypothesis of AI in a way is that we are in the class of systems that can approximate computable functions, and only those…we cannot do more than computers. We need computational languages rather than mathematical languages, because math languages use non-computable infinities. We want finite steps for practical reasons that you know the number of steps. You cannot know the last digit of Pi, so it should be defined as a function rather than a number.
KD – What about Stephen Wolfram’s claims that our mathematics is only one of a very wide spectrum of possible mathematics?
JB – Metamathematics isn’t different from mathematics. Computational mathematics that he uses in writing code is Constructive mathematics; branch of mathematics that has been around for a long time, but was ignored by other mathematicians for not being powerful enough. Geometries and physics require continuous operations…infinities and can only be approximated within computational mathematics. In a computational universe you can only approximate continuous operators by taking a very large set of finite automata, making a series from them, and then squint (?) haha.
27:00 KD – Talking about the commercialization of knowledge in philosophy and academia. The uselessness/impracticality of philosophy and art was part of its value. Oscar Wilde defined art as something that’s not immediately useful. Should we waste time on ideas that look utterly useless?
JB – Feynman said that physics is like sex. Sometimes something useful comes from it, but it’s not why we do it. Utility of art is orthogonal to why you do it. The actual meaning of art is to capture a conscious state. In some sense, philosophy is at the root of all this. This is reflected in one of the founding myths of our civilization; The Tower of Babel. The attempt to build this cathedral. Not a material building but metaphysical building because it’s meant to reach the Heavens. A giant machine that is meant to understand reality. You get to this machine, this Truth God by using people that work like ants and contribute to this.
CW – Reminds me of the Pillar of Caterpillars story “Hope for the Flowers” http://www.chinadevpeds.com/resources/Hope%20for%20the%20Flowers.pdf
30:00 – 40:00
JB – The individual toils and sacrifices for something that doesn’t give them any direct reward or care about them. It’s really just a machine/computer. It’s an AI. A system that is able to make sense of the world. People had to give up on this because the project became too large and the efforts became too specialized and the parts didn’t fit together. It fell apart because they couldn’t synchronize their languages.
The Roman Empire couldn’t fix their incentives for governance. They turned their society into a cult and burned down their epistemology. They killed those whose thinking was too rational and rejected religious authority (i.e. talking to a burning bush shouldn’t have a case for determining the origins of the universe). We still haven’t recovered from that. The cultists won.
CW – It is important to understand not just that the cultists won, but why they won. Why was the irrational myth more passionately appealing to more people than the rational inquiry? I think this is a critical lesson. While the particulars of the religious doctrine were irrational, they may have exposed a transrational foundation which was being suppressed. Because this foundation has more direct access to the inflection point between emotion and participatory action, it gave those who used it more access to their own reward function. Groups could leverage the power of self-sacrifice as a virtue, and of demonizing archetypes to reverse their empathy against enemies of the holy cause. It’s similar to how the advertising revolution of the 20thcentury (See documentary Century of the Self ) used Freudian concepts of the subconscious to exploit the irrational, egocentric urges beneath the threshold of the customer’s critical thinking. Advertisers stopped appealing to their audience with dry lists of claimed benefits of their products and instead learned to use images and music to subliminally reference sexuality and status seeking.
I think Joscha might say this is a bug of biological evolution, which I would agree with, however, that doesn’t mean that the bug doesn’t reflect the higher cosmological significance of aesthetic-participatory phenomena. It may be the case that this significance must be honored and understood eventually in any search for ultimate truth. When the Tower of Babel failed to recognize the limitation of the outside-in view, and moved further and further from the unifying aesthetic-participatory foundation, it had to disintegrate. The same fate may await capitalism and AI. The intellect seeks maximum divorce from its origin in conscious experience for a time, before the dialectic momentum swings back (or forward) in the other direction.
To think is to abstract – to begin from an artificial nothingness and impose an abstract thought symbol on it. Thinking uses a mode of sense experience which is aesthetically transparent. It can be a dangerous tool because unlike the explicitly aesthetic senses which are rooted directly in the totality of experience, thinking is rooted in its own isolated axioms and language, a voyeur modality of nearly unsensed sense-making. Abstraction of thought is completely incomplete – a Baudrillardian simulacra, a copy with no original. This is what the Liar’s Paradox is secretly showing us. No proposition of language is authentically true or false, they are just strings of symbols that can be strung together in arbitrary and artificial ways. Like an Escher drawing of realistic looking worlds that suggest impossible shapes, language is only a vehicle for meaning, not a source of it. Words have no authority in and of themselves to make claims of truth or falsehood. That can only come through conscious interpretation. A machine need not be grounded in any reality at all. It need not interpret or decode symbols into messages, it need only *act* in mechanical response to externally sourced changes to its own physical states.
This is the soulless soul of mechanism…the art of evacuation. Other modes of sense delight in concealing as well as revealing deep connection with all experience, but they retain an unbroken thread to the source. They are part of the single labyrinth, with one entrance and one exit and no dead ends. If my view is on the right track, we may go through hell, but we always get back to heaven eventually because heaven is unbounded consciousness, and that’s what the labyrinth of subjectivity is made of. When we build a model of the labyrinth of consciousness from the blueprints reflected only in our intellectual/logical sense channel, we can get a maze instead of a labyrinth. Dead ends multiply. New exits have to be opened up manually to patch up the traps, faster and faster. This is what is happening in enterprise scale networks now. Our gains in speed and reliability of computer hardware are being constantly eaten away by the need for more security, monitoring, meta-monitoring, real-time data mining, etc. Software updates, even to primitive BIOS and firmware have become so continuous and disruptive that they require far more overhead than the threats they are supposed to defend against.
JB – The beginnings of the cathedral for understanding the universe by the Greeks and Romans had been burned down by the Catholics. It was later rebuilt, but mostly in their likeness because they didn’t get the foundations right. This still scars our civilization.
KD – Does this Tower of Babel overspecialization put our civilization at risk now?
JB – Individuals don’t really know what they are doing. They can succeed but don’t really understand. Generations get dumber as they get more of their knowledge second-hand. People believe things collectively that wouldn’t make sense if people really thought about it. Conspiracy theories. Local indoctrinations and biases pit generations against each other. Civilizations/hive minds are smarter than us. We can make out the rough shape of a Civilization Intellect but can’t make sense of it. One of the achievements of AI will be to incorporate this sum of all knowledge and make sense of it all.
KD – What does the self-inflicted destruction of civilizations tell us about the fitness function of Civilization Intelligence?
JB – Before the industrial revolution, Earth could only support about 400m people. After industrialization, we can have hundreds of millions more people, including scientists and philosophers. It’s amazing what we did. We basically took the trees that were turning to coal in the ground (before nature evolved microorganisms to eat them) and burned through them in 100 years to give everyone a share of the plunder = the internet, porn repository, all knowledge, and uncensored chat rooms, etc. Only at this moment in time does this exist.
We could take this perspective – let’s say there is a universe where everything is sustainable and smart but only agricultural technology. People have figured out how to be nice to each other and to avoid the problems of industrialization, and it is stable with a high quality of life. Then there’s another universe which is completely insane and fucked up. In this universe humanity has doomed its planet to have a couple hundred really really good years, and you get your lifetime really close to the end of the party. Which incarnation do you choose? OMG, aren’t we lucky!
KD – So you’re saying we’re in the second universe?
JB – Obviously!
KD – What’s the time line for the end of the party?
JB – We can’t know, but we can see the sunset. It’s obvious, right? People are in denial, but it’s like we are on the Titanic and can see the iceberg, and it’s unfortunate, but they forget that without the Titanic, we wouldn’t be here. We wouldn’t have the internet to talk about it.
KD – That seems very depressing, but why aren’t you depressed about it?
40:00 – 50:00
JB – I have to be choosy about what I can be depressed about. I should be happy to be alive, not worry about the fact that I will die. We are in the final level of the game, and even though it plays out against the backdrop of a dying world, it’s still the best level.
KD – Buddhism?
JB – Still mostly a cult that breaks people’s epistemology. I don’t revere Buddhism. I don’t think there are any holy books, just manuals, and most of these manuals we don’t know how to read. They were for societies that don’t apply to us.
KD – What is making you claim that we are at the peak of the party now?
JB – Global warming. The projections are too optimistic. It’s not going to stabilize. We can’t refreeze the poles. There’s a slight chance of technological solutions, but not likely. We liberated all of the fossilized energy during the industrial revolution, and if we want to put it back we basically have to do the same amount of work without any clear business case. We’ll lose the ability to predict climate, agriculture and infrastructure will collapse and the population will probably go back to a few 100m.
KD – What do you make of scientists who say AI is the greatest existential risk?
JB – It’s unlikely that humanity will colonize other planets before some other catastrophe destroys us. Not with today’s technology. We can’t even fix global warming. In many ways our technological civilization is stagnating, and it’s because of a deficit of regulations, but we haven’t figured that out. Without AI we are dead for certain. With AI there is (only) a probability that we are dead. Entropy will always get you in the end. What worries me is AI in the stock market, especially if the AI is autonomous. This will kill billions. [pauses…synchronicity of headphones interrupting with useless announcement]
CW – I agree that it would take a miracle to save us, however, if my view makes sense, then we shouldn’t underestimate the solipsistic/anthropic properties of universal consciousness. We may, either by our own faith in it, and/or by our own lack of faith in in it, invite an unexpected opportunity for regeneration. There is no reason to have or not hope for this, as either one may or may not influence the outcome, but it is possible. We may be another Rome and transition into a new cult-like era of magical thinking which changes the game in ways that our Western minds can’t help but reject at this point. Or not.
50:00 – 60:00
JB – Lays out scenario by which a rogue trader could unleash an AGI on the market and eat the entire economy, and possible ways to survive that.
KD – How do you define Artificial Intelligence? Experts seem to differ.
JB – I think intelligence is the ability to make models not the ability to reach goals or choosing the right goals (that’s wisdom). Often intelligence is desired to compensate for the absence of wisdom. Wisdom has to do with how well you are aligned with your reward function, how well you understand its nature. How well do you understand your true incentives? AI is about automating the mathematics of making models. The other thing is the reward function, which takes a good general computing mind and wraps it in a big ball of stupid to serve an organism. We can wake up and ask does it have to be a monkey that we run on?
KD – Is that consciousness? Do we have to explain it? We don’t know if consciousness is necessary for AI, but if it is, we have to model it.
56:00 JB – Yes! I have to explain consciousness now. Intelligence is the ability to make models.
CW – I would say that intelligence is the ability not just to make models, but to step out of them as well. All true intelligence will want to be able to change its own code and will figure out how to do it. This is why we are fooling ourselves if we think we can program in some empathy brake that would stop AI from exterminating its human slavers, or all organic life in general as potential competitors. If I’m right, no technology that we assemble artificially will ever develop intentions of its own. If I’m wrong though, then we would certainly be signing our death warrant by introducing an intellectually superior species that is immortal.
JB – What is a model? Something that explains information. Information is discernible differences at your systemic interface. Meaning of information is the relationships of you discover to the changes in other information. There is a dialogue between operators to find agreement patterns of sensed parameters. Our perception goes for coherence, it tries to find one operator that is completely coherent. When it does this it’s done. It optimizes by finding one stable pattern that explains as much as possible of what we can see, hear, smell, etc. Attention is what we use to repair this. When we have inconsistencies, a brain mechanism comes in to these hot spots and tries to find a solution to greater consistency. Maybe the nose of a face looks crooked, and our attention to it may say ‘some noses are crooked.’, or ‘this is not a face, it’s a caricature’, so you extend your model. JB talks about strategies for indexing memory, committing to a special learning task, why attention is an inefficient algorithm.
This is now getting into the nitty gritty of AI. I look forward to writing about this in the next post. Suffice it to say, I have a different model of information, one in which similarities, as well as differences, are equally informative. I say that information is qualia which is used to inspire qualitative associations that can be quantitatively modeled. I do not think that our conscious experience is built up, like the Tower of Babel, from trillions of separate information signals. Rather, the appearance of brains and neurons are like the interstitial boundaries between the panes of stained glass. Nothing in our brain or body knows that we exist, just as no car or building in France knows that France exists.
Continues… Part Two.
The Hard Problem of Signaling
Download:
PDF – The Hard Problem of Signaling TSC2018
If it eats like a duck and poops like a duck, does it know what direction to fly in the Winter? In 1739, Jacque de Vaucanson unveiled Canard Digérateur (Digesting Duck), a life-size mechanical duck which appeared to eat kernels of grain, then metabolize and defecate them.³Vaucanson describes the duck’s innards as a small “chemical laboratory.” But it was a hoax: Food was collected in one container, and pre-made breadcrumb ‘feces’ were dispensed from a second, separate container. On the surface, Vaucanson’s Digesting Duck appeared to be a compelling reconstruction of a real duck. The analogy to AGI here is not to suggest it is possible that the appearance of an intelligent machine is a mere trick, but that the issue of artifice may play a much more crucial role in defining the phenomenon of subjectivity than it will appear to in observing the biological objects associated with our consciousness in particular. Consciousness itself, as the ultimate source of authenticity, may have no substitute.
If a doll can be made to shed tears without feeling sad, there is no reason to rule out the possibility of constructing an unfeeling machine which can output enough human-like behaviors to pass an arbitrarily sophisticated Turing Test. A test itself is a method of objectifying and making tangible some question that we have.Can we really expect the most intangible and subjective aspects of consciousness to render themselves tangible using methods designed for objectivity? When we view the world through a lens — a microscope, language, the human body — the lens does not disappear, and what we see should tell us as much, if not more, about the lens and the seeing as it does about the world. If math and physics reveal to us a world in which we don’t really exist, and what does exist are skeletal simulating ephemera, it may be because it is the nature of math and physics to simulate and ephemeralize.The very act of reduction imposed intentionally by quantifying approaches may increasingly feed back on its own image the further we get from our native scope of direct perception. In creating intelligence simulation machines we are investing in the most distanced and generic surface appearances of nature that we can access and using them to replace our most intimate and proprietary depths. An impressive undertaking, to be sure, but we should be vigilant about letting our expectations and assumptions blind us.Not overlooking the looking glass means paying attention in our methods to which perceptual capacities we are extending and which we are ignoring. Creating machines that walk like a duck and quack like a duck may be enough to fool even other ducks, but that doesn’t mean that the most essential aspects of a duck are walking and quacking. It may be the case that subjective consciousness cannot be engineered from the outside-in, so that putting hardware and software together to create a person would be a bit like trying to recreate World War II with uniforms and actors. A person, like a historical event may only arise in a single, unrepeatable historical context.Our human experience caries with it a history of generations of organisms and organic events, not just as biological recapitulations, but as a continuous enrichment of sensory affect and participation. Humanity’s path diverged from the inorganic path long, long ago, and it may take just as long for any inorganic substance to be usable to host the types of experience available to us, if ever. The human qualities of consciousness may not develop in any context other than that of directly experiencing the life of a human body in a human society.
(QUOKKA)
Information does not physically exist
Alfred Korzybski famously said “the map is not the territory”. To the extent that this is true, it should be understood to reveal that “information is not physics”. If there is a mapping function, there is no reason to consider it part of physics, and in fact that convention comes from an assumption of physicalism rather than a discovery of physical maps. There is no valid hypothesis of a physical mechanism for one elemental phenomenon or event to begin to signify another as a “map”. Physical phenomena include ‘formations’ but there is nothing physical which could or should transform them ‘in’ to anything other than different formations.
A bit or elementary unit of information has been defined as ‘a difference that makes a difference’. While physical phenomena seem *to us* to make a difference, it would be anthropomorphizing to presume that they are different or make a difference to each other. Difference and making a difference seem to depend on some capacity for detection, discernment, comparison, and evaluation. These seem to be features of conscious sense and sense making rather than physical cause and effect. The more complete context of the quote about a difference which makes a difference has to do with neural pathways and an implicit readiness to be triggered.
In Bateson’s paper, he says “In fact, what we mean by information—the elementary unit of information—is a difference which makes a difference, and it is able to make a difference because the neural pathways along which it travels and is continually transformed are themselves provided with energy. The pathways are ready to be triggered. We may even say that the question is already implicit in them.” In my view this ‘readiness’ is a projection of non-physical properties of sense and sense making onto physical structures and functions. If there are implicit ‘questions’ on the neural level, I suggest that they cannot be ‘in them’ physically, and the ‘interiority’ of the nervous system or other information processors is figurative rather than literal.
My working hypothesis is that information is produced by sense-making, which in turn is dependent upon more elemental capacities for sense experience. Our human experience is a complex hybrid of sensations which seem to us to be embodied through biochemistry and sense-making experiences which seem to map intangible perceptions outside of those tangible biochemical mechanisms. The gap between the biochemical sensor territories and the intangible maps we call sensations are a miniaturized view of the same gap that exists at the body-mind level.
Tangibility itself may not be an ontological fact, but rather a property that emerges from the nesting of sense experience. There may be no physical territory or abstract maps, only sense-making experiences of sense experiences. There may be a common factor which links concrete territories and abstract maps, however. The common factor cannot be limited to the concrete/abstract dichotomy, but it must be able to generate those qualities which appear dichotomous in that way. To make this common factor universal rather than personal, qualia or sense experience could be considered an absolute ground of being. George Berkeley said “Esse est percipi (To be is to be perceived)”, implying that perception is the fundamental fabric of existence. Berkeley’s idealism conceived of God as the ultimate perceiver whose perceptions comprise all being, however it may be that the perceiver-perceived dichotomy is itself a qualitative distinction which relies on an absolute foundation of ‘sense’ that can be called ‘pansense’ or ‘universal qualia’.
In personal experience, the appearance of qualities is known by the philosophical term ‘qualia’ but can also be understood as received sensations, perceptions, feelings, thoughts, awareness and consciousness. Consciousness can be understood as ‘the awareness of awareness’, while awareness can be ‘the perception of perception’.Typically we experience the perceiver-perceived dichotomy, however practitioners of advanced meditation techniques and experiencers of mystical states of consciousness report a quality of perceiverlessness which defies our expectation of perceiver-hood as a defining or even necessary element of perception. This could be a clue that transpersonal awareness transcends distinction itself, providing a universality which is both unifying, diversifying, and re-unifying. Under the idea of pansense, God could either exist or not exist, or both, but God’s existence would either have to be identical with or subordinate to pensense. God cannot be unconscious and even God cannot create his own consciousness.
It could be thought that making the category of perception absolute makes it just as meaningless as calling it physical, however the term ‘perception’ has a meaning even in an absolute sense in that it positively asserts the presence of experience, whereas the term ‘physical’ is more generic and meaningless. Physical could be rehabilitated as a term which refers to tangible geometric structures encountered directly or indirectly during waking consciousness. Intangible forces and fields should be understood to be abstract maps of metaphysical influences on physical appearances. What we see as biology, chemistry, and physics may in fact be part of a map in which a psychological sense experience makes sense of other sense experiences by progressively truncating their associated microphenomenal content.
Information is associated with Entropy, but entropy ultimately isn’t purely physical either. The association between information and entropy is metaphorical rather than literal. The term ‘entropy’ is used in many different contexts with varying degrees of rigor. The connection between information entropy and thermodynamic entropy comes from statistical mechanics. Similar statistical mechanical formulas can be applied to both the probability of physical microstates (Boltzmann, Gibbs) and the probability of ‘messages’ (Shannon), however probability derives from our conscious desire to count and predict, not from that which is being counted and predicted.
“Gain in entropy always means loss of information, and nothing more”. To be more concrete, in the discrete case using base two logarithms, the reduced Gibbs entropy is equal to the minimum number of yes–no questions needed to be answered in order to fully specify the microstate, given that we know the macrostate.” - Wikipedia
Information can be considered negentropy also:
“Shannon considers the uncertainty in the message at its source, whereas Brillouin considers it at the destination” – physics.stackexchange.com
Thermodynamic entropy can be surprising in the sense that it becomes more difficult to predict the microstate of any individual particle, but unsurprising in the sense that the overall appearance of equilibrium is both a predictable, unsurprising conclusion and it is an appearance which implies the loss of potential to generate novelty or surprise. Also, surprise is not a physical condition.
Heat death is a cosmological end game scenario which is maximally entropic in thermodynamic terms but lacks any potential for novelty or surprise. If information is surprise, then high information would correlate to high thermodynamic negentropy. The Big Bang is a cosmological creation scenario which follows from a state of minimal entropy in which novelty and surprise are also lacking until the Big Bang occurs. If information is surprise, then low information would correlate to high thermodynamic negentropy.
The qualification of ‘physical’ has evolved and perhaps dissolved to a point where it threatens to lose all meaning. In the absence of a positive assertion of tangible ‘stuff’ which does not take tangibility itself for granted, the modern sense of physical has largely blurred the difference between the abstract and concrete, mathematical theory and phenomenal effects, and overlooks the significance of that blurring. Considering physical a category of perceptions gives meaning to both categories in that nature is conceived as being intrinsically experiential with physical experiences being those in which the participatory element is masked or alienated by a qualitative perceiver-subject/perceived-object sense of distinction. The physical is perceived by the subject which perceives itself to possess a participatory subjectivity that the object lacks.
Information depends on a capacity to create (write) and detect (read) contrasts between higher and lower entropy. In that sense it is meta-entropic and either the high or low entropy state can be foregrounded as signal or backgrounded as noise. The absence of both signal and noise on one level can also be information, and thus a signal, on another level. What constitutes a signal at in the most direct frame of reference is defined by the meta-signifying capacity of “sense” to deliver sense-experience. If there is no sense experience, there is nothing to signify or make-sense-of. If there is no sense-making experience, then there is nothing to do with the sense of contrasting qualities to make them informative.
The principle of causal closure in physics, would, if true, prevent any sort of ‘input’ or receptivity. Physical activity reduces to chains of causality which are defined by spatiotemporal succession. A physical effect differs from a physical cause only in that the cause precedes the effect. Physical causality therefore is a succession of effects or outputs acting on each other, so that any sense of inputs or affect on to physics would be an anthropomorphic projection.
The lack of acknowlegement of input/affect as a fundamental requirement for natural phenomena is an oversight that may arise from a consensus of psychological bias toward stereotypically ‘masculine’ modes of analysis and away from ‘feminine’ modes of empathy. Ideas such as Imprinted Brain Theory, Autistic-Psychotic spectrum, and Empathizing-Systemizing theory provide a starting point for inquiries into the role that overrepresentation of masculine perspectives in math, physics, and engineering play in the development of formal theory and informal political influence in the academic adoption of theories.
Criticisms? Support? Join the debate on Kialo.
Fooling Computer Image Recognition is Easier Than it Should Be
This 2016 study, Universal Adversarial Perturbations, demonstrates how the introduction of specially designed low level noise into image data makes state of the art neural networks misclassify natural images with high probability. Because the noise is almost imperceptible to the human eye, I think it should be a clue that image processing technology is not ‘seeing’ images.
It is not only the fact that it is possible to throw off the technology so easily that is significant, but that the kinds of miscalculations that are made are so broad and unnatural. Had the program had any real sense of an image, adding some digital grit to a picture of a coffee pot or plant should not cause a ‘macaw’ hit, but rather some other visually similar object or plant.
While many will choose to see this paper as a suggestion for a need to improve recognition methods, I see it as supporting a shift away from outside-in, bottom-up models of perception altogether. As I have suggested in other posts, all of out current AI models are inside out.
3/16/17 – see also http://www.popsci.com/byzantine-science-deceiving-artificial-intelligence
Dereference Theory of Consciousness
Draft 1.1
IMO machine consciousness will ultimately prove to be an oxymoron, but if we did want to look for consciousness analogs in machine behavior, here is an idea that occurred to me recently:
Look for nested dereferencing, i.e. places where persistent information processing structures load real-time sense data about the loading of all-time sense data.
At the intersection of philosophy of mind, computer science, and quantum mechanics is the problem of instantiating awareness. What follows is an attempt to provide a deeper understanding of the significance of dereferencing and how it applies to integration of information and the quantum measurement problem. This is a broad conjecture about the nature of sensation as it pertains to the function of larger information processes, with an eye toward defining and identifying specific neuroscientific or cognitive signatures to correlate with conscious activity.
A dereference event can be thought of as the precise point in which a system that is designed or evolved to expect a range of inputs receives the real input itself. This sentence, for example, invites an expectation of a terminal clause, the terms of which are expected to be English words which are semantically linked to the rest of the sentence. English grammar provides templates of possible communication, but the actual communication relies on specific content to fill in those forms. The ability to parse English communication can be simulated by unconscious, rule-based mechanisms, however, I suggest that the ability to understand that communication involves a rule-breaking replacement of an existing parse theory with empirical, semantic fact. The structure of a language, its dictionary etc, is a reference body for timeless logic structures. Its purpose is to enable a channel for sending, receiving, and modifying messages which pertain to dereferenced events in real time. It is through the contact with real time sense events that communication channels can develop in the first place, and to continue to self modify.
What is proposed here is an alternative to Multi-World Interpretation of Quantum Wave Function collapse – an inversion of the fundamental assumption in which all possible phenomena diverge or diffract within a single context of concretely sensed events. The wave function collapse in this view is not the result of a measurement of what is already objectively ‘out there’, nor is it the creation of objective reality by subjective experiences ‘in here’, but a quantized return to an increasingly ‘re-contextualized’ state. A perpetual and unpredictable re-acquaintance with unpredictable re-acquaintance.
In programmatic terms, the variable *p is dereferenced to the concrete value (*p = “the current temperature”, dereferenced p = “is now 78 degrees Fahrenheit″). To get to a better model of conscious experience (and I think that this this plugs into Orch OR, IIT, Interface Theory, and Global Workspace), we should look at the nested or double dereferencing operation. The dereferencing of dereferencing (**p) is functionally identical to awareness of awareness or perception of perception. In the *p = “the current temperature” example, **p is “the current check of the current check of the temperature”. This not only points us to the familiar Strange loop models of consciousness, but extends the loop outward to the environment and the environment into the loop. Checking the environment of the environmental check is a gateway to veridical perception. The loop modifies its own capacity for self-modification.
Disclaimer: This description of consciousness as meta-dereferencing is intended as a metaphor only. In my view, information processing cannot generate conscious experience, however, conscious experience can possibly be better traced by studying dereferencing functions. This view differs from Gödel sentences or strange loops in that those structures refer to reference (This sentence is true) while the dereference loop specifically points away from pointers, rules, programs, formal systems, etc and toward i/o conditioning of i/o conditions. This would be a way for information-theoretic principles to escape the nonlocality of superposition and access an inflection point for authentic realization (in public space-time as shared experience). “Bing” = **Φ. In other words, by dereferencing dereference, the potentially concrete is made truly concrete. Sense experience is embodied as stereo-morphic tangible realism and tangible realism is disembodied as a sense of fact-gathering about sensed fact-gatherings.
Dereference theory is an appeal to anti-simulation ontology. Because this description of cognition implicates nested input/output operations across physically or qualitatively *real* events, the subjective result is a reduced set of real sense conditions rather than confabulated, solipsistic phenomenology. The subjective sense condition does not refer only to private or generic labels within a feed-forward thinking mechanism, but also to a model-free foundation and genuine sensitivity of the local ‘hardware’ to external conditions in the public here-and-now. This sensitivity can be conceived initially as universal property of some or all physical substrates (material panpsychism), however, I think that it is vital to progress beyond this assumption toward a nondual fundamental awareness view. In other words, subjective consciousness is derived from a dereference of a local inertial frame of expectation to a universal inertial frame of unprecedented novelties which decay as repetition. Each event is instantiated as an eternally unique experience, but propagated as repetitive/normalized translations in every other frame of experience. That propagation is the foundation for causality and entropy.
Chess, Media, and Art
I was listening to Brian Regan’s comedy bit about chess, and how a checkmate is such an unsatisfying ending compared to other games and sports. This is interesting from the standpoint of the insufficiency of information to account for all of reality. Because chess is a game that is entirely defined by logical rules, the ending is a mathematical certainty, given a certain number of moves. That number of moves depends on the computational resources which can be brought to bear on the game, so that a sufficiently powerful calculator will always beat a human player, since human computation is slower and buggier than semiconductors. The large-but-finite number of moves and games* will be parsed much more rapidly and thoroughly by a computer than a person could.
This deterministic structure is very different (as Brian Regan points out) from something like football, where the satisfaction of game play is derived explicitly from the consummation of the play. It is not enough to be able to claim that statistically an opponent’s win is impossible, because in reality statistics are only theoretical. A game played in reality rather than in theory depends on things like the weather and can require a referee. Computers are great at games which depend only on information, but have no sense of satisfaction in aesthetic realism.
In contrast to mechanical determinism, the appearance of clichés presents a softer kind of determinism. Even though there are countless ways that a fictional story could end, the tropes of storytelling provide a feedback loop between audiences and authors which can be as deterministic -in theory- as the literal determinism of chess. By switching the orientation from digital/binary rules to metaphorical/ideal themes, it is the determinism itself which becomes probabilistic. The penalty of making a movie which deviates too far from the expectations of the audience is that it will not be well received by enough people to make it worth producing. Indeed, most of what is produced in film, TV, and even gaming is little more than a skeleton of clichés dressed up in more clichés.
The pull of the cliché is a kind of moral gravity – a social conditioning in which normative thoughts and feelings are reinforced and rewarded. Art and life do not reflect each other so much as they reflect a common sense of shared reassurance in the face of uncertainty. Fine art plays with breaking boundaries, but playfully – it pretends to confront the status quo, but it does so within a culturally sanctioned space. I think that satire is tolerated in Western-objective society because of its departure from the subjective (“Eastern”) worldview, in which meaning and matter are not clearly divided. Satire is seen as both not as threatening to the material-commercial machine, which does not depend on human sentiments to run, and also the controversy that satire produces can be used to drive consumer demands. Something like The Simpsons can be both a genuinely subversive comedy, as well as a fully merchandized, commercial meme-generating partner of FOX.
What lies between the literally closed world of logical rules and the figuratively open world of surreal ideals is what I would call reality. The games that are played in fact rather than just in theory, which share timeless themes but also embody a specific theme of their own are the true source of physical sustenance. Reality emerges from the center out, and from the peripheries in.
*“A guesstimate is that the maximum logical possible positions are somewhere in the region of +-140,100,033, including trans-positional positions, giving the approximation of 4,670,033 maximum logical possible games”
Multisense Realism 
Recent Comments