Goodbye Mr Searle

I've been interested in artificial intelligence since I first got into computers - my teens. Back then there was a lot more hype about the potential of AI and a break through was always "just around the corner". Twenty years on no real breakthroughs have arrived and autonomous robots are still pretty primitive, the most compelling being those that emulate an insect without any real "intelligence" as we'd know it.

From the books on AI I read back then there was always one loud-mouthed naysayer claiming that computers could never be intelligent - John Searle. His standard argument against AI was his "Chinese room" which never made any sense to me and as best I can tell a huge number of other more learned people. But on every occasion he trots it out like a broken record and asserts that its a QED for strong AI being impossible.

Along with the Chinese room argument Searle would also talk about how intelligence is a phenomenon that only organic "wet brain" systems could have. This was later finessed to saying that the brain does something magical on a quantum level and since computers are just switches they could never "think". All along I've suspected that Searle is actually a religious person who has some deep seated belief that only God's critters could think and that for man to create its own thinking critters would be unthinkable. I don't know if this is true but his passion and monotony of argument on this point without much substance has always had that tone of a religious conviction.

I've never been at all convinced by Searle's "Chinese room" argument (I wont expand on it here, just look it up) since if it was true then it could be equally applied to human brains i.e. a brain can't possibly exhibit intelligence because the neurons don't understand the symbols they process. So if a brain made up of neurons can be intelligent then why not a brain made up of something else, say man made "switches" be intelligent?

Then again Searle's impassioned appeal to requirement of "wet brain" using quantum level effects being necessary (think parallel processing, fuzzy logic, uncertainty principle) for intelligence had at least some appeal to me. Not because I think its a unique "web brain" thing, but that a "wet
brain" may possess features that yield sufficient computing power to realize intelligence as we'd know it. You see I'm a believer that intelligence requires self awareness i.e. consciousness, and that there are levels of consciousness and hence levels of intelligence that critters exhibit.

By my theory humans are probably the most self-aware of critters on the planet, other mammals coming close seconds, with others being somewhat less self aware. A bug probably has next to no self awareness, a bird significantly more, whereas bacteria, amoeba and the like are a zero. I also think that a human gains self awareness and intelligence as it grows - clearly it starts at zero when an embryo, and develops to a much higher level by birth and 100% by a few years.

So I think that a computer could be made self aware by programming alone, but the amount of computation required to emulate that of a human is many orders of magnitude beyond that possible today. Hence the best we can do is emulate the intelligence of a bug. To some extent the problem may be that human brains have huge numbers of neurons and they all do stuff in parallel and the synergistic effect of all the parallel computing is way more than could reasonably computed in real-time with largely sequential computation. Maybe network all the computers on the planet together and you could start to get somewhere, perhaps a bird brain vs. a bug-brain. But in principle I think it would be possible to program and build a system with a high enough level of self awareness and intelligence and that intelligence is not a pure wet brain only phenomenon.

Anyway I am pleased to read that a quantum level computing device may soon be available that can compute many different things in parallel. This is not the same as having multiple core conventional processors working in parallel. So the "16 cubits" the new processor has does not simply yield sixteen times as much computing power as a 16-core processor CPU from Intel might (if performing optimally) but actually 2 to the power of 16 times the power, i.e. 64,000 times or so. Expanding this if you had a 32 cubit system you could get 4 billion times the processing power, and a 64 cubit system would give you 16 quintillion (16 billion billion) times as much computational power. At these levels I am firmly convinced we will see self-aware artificially intelligent systems that approach the level of intelligence observed in mammals and humans.

No, I have no idea of the programming model or programs themselves required to emulate the mammalian brain, but I think at least in principle the advent of these new quantum level computing devices decisively removes any "wet brain" vs. silicon brain barriers to entry in the intelligence club. Since I never had any truck with Searle's Chinese room "argument" I can therefore conclude that all of Searle's objections to computers exhibiting strong AI are now eradicated. Goodbye Mr Searle!