|
Lost in Hilbert Spice
Join Date: Jun 2004
Location: Surrounded by knaves and fools
Posts: 3,506
|
SPEAKERS:
Lord Martin Rees, Astronomer Royal
Jaan Tallinn, co-founder of Skype
Huw Price, Bertrand Russell Professor of Philosophy at Cambridge
And some more David Pearce on digital beings
Quote:
[on whether digital zombies can investigate sentience]
Neurons are indeed fabulously complex information processors. Thus e.g. the different amino acid sequences and secondary, tertiary and quaternary protein folding structures internal to the neuron may well be implicated in innumerable different microqualia. But once again, I'm at a loss to know how a digital computer could investigate the first-person/third person psychophysical mapping needed to understand their relationship. Even with the master equation of a formally complete Theory of Everything, you'll need to instantiate some of its solutions to understand them. "Mary" (cf. Knowledge argument - Wikipedia, the free encyclopedia) - or a digital computer - never will understand redness. More to the point, ignorance of the phenomenal nature of pain and pleasure entails a digital computer will never understand why anything matters at all. Insidiously, the Church-Turing thesis has promoted an impoverished conception of what constitutes a well-defined problem an intelligent agent can investigate.
[...]We are certainly acutely aware of some of our logical stumbles. But I'd argue that one is most vividly aware of an evolutionarily ancient process that works extraordinarily well - and is completely beyond any digital computer, which is "not even stupid". The classical world is an artefact of quantum minds. What one naively calls "perceiving one's surroundings" actually entails generating "bound" and cross-modally matched experiential objects in a unitary world-simulation run by a (fleetingly) unitary self - and in almost real time to boot.
* * *
Can a cognitive agent be intelligent, let alone superintelligent, and yet fail to understand, or lack any capacity to investigate, fundamental features of the natural world? If the agent in question were constitutionally ignorant of the properties of, say, matter, energy and the second law of thermodynamics, then we would say no: such an agent is profoundly ignorant, or at best an idiot savant. Yet if the cognitive agent in question is constitutionally ignorant of the properties of, say, phenomenal objects, conscious minds, or the nature of pain and pleasure, then many AI researchers are nonetheless willing to ascribe intelligence - and potentially even superintelligence. IMO this is an anthropomorphic projection on our part.
How might the apologist for digital (super)intelligence respond? Several ways, I guess. Here are just two.
First, s/he might argue that the manifold varieties of consciousness are unimportant and/or causally impotent. Intelligence, and certainly not superintelligence, does not concern itself with trivia.
But in what sense are, say, the experience of agony or despair trivial, whether subjectively to their victim, or conceived as disclosing a feature of the natural world? Compare how, in a notional zombie world otherwise physically type-identical to our world, nothing would inherently matter at all. Some of our supposed counterparts might undergo boiling in oil, but who cares: they aren't sentient. By contrast to such a fanciful zombie world, the nature of phenomenal agonies as we undergo such states isn't trivial: indeed the thought that (1) I'm in unbearable agony and (2) the agony doesn't matter, is devoid of cognitive sense. And in any case, we can be sure that phenomenal properties aren't causally impotent epiphenomena. Epiphenomena, by definition, lack causal efficacy - and hence lack the ability physically or functionally to stir us to write and talk about their existence.
Second, the believer in digital (super)intelligence might claim that (some of the programs executed by) digital computers are conscious, or at least potentially conscious, not least future software emulations of human brains. For reasons we admittedly don't yet understand, some physical states of matter and energy, perhaps the different algorithms executed in various information processors, are identical with different states of consciousness, i.e. a functionalist version of the mind-brain identity theory is correct. Granted, we don't yet understand the mechanisms by which information processing generates consciousness. But whatever these consciousness-generating processes may turn out to be, materialism is correct. Biological and nonbiological agents alike can be conscious minds.
Unfortunately, there is an insurmountable problem here. Identity is not a causal relationship. We can't simultaneously claim that a conscious state is identical with a brain state and maintain that this brain state causes (or "generates", or "gives rise to" etc) the conscious state in question. Nor - and this is where Searle stumbles - can causality operate between what are only levels of description. Hence the Hard Problem of Consciousness and the Explanatory Gap. What I meant by denying that consciousness is a mere puzzle is that the solution to puzzles don't challenge our conceptual scheme. Thus a difficult crossword clue may stump us; but we may be confident the answer will leave our world-picture intact. By contrast, if we discovered a fairy living at the bottom of the garden, even a little one, then materialism would be falsified. Materialism is the thesis that the physical facts exhaustively constitute all the facts. The existence of consciousness - even a single instance of consciousness - falsifies materialism. Actually, not everyone would agree here. Radical eliminativism about consciousness has been described as the craziest theory in the history of philosophy; but eliminativists are right in one sense: given the ontology of physics as standardly understood, consciousness is impossible. Most of us find eliminativism literally incredible.
Anyhow, the conjecture I offer to resolve the mystery of conscious mind, involving a combination of Strawsonian physicalism plus macroscopic quantum coherence, may most likely be false. But it's empirically adequate, eliminates the Hard Problem and the Explanatory Gap, and predicts that digital computers will never be sentient. We shall see. :-)
[...]Dustin, I guess we differ on whether or not our computers show abstractions can or can't have causal efficacy. Pragmatically, of course, it's hugely useful to pretend they do - just as pragmatically it's useful to think of the lump of silicon in front of me in terms of the different abstraction layers of a computer architecture (i.e. hardware, firmware, assembler, kernel, operating system and applications). But everything that takes place in your PC supervenes on microphysical interactions whose behaviour is exhaustively described by the physics of the Standard Model (or its ultimate successor). Reality only has one "level" - and that's where all the work gets done. As you know, I take an equally reductive approach to consciousness.
|
|