CoyoteNature wrote:I think that in general the AI crowd tend to be overly optimistic about its prospects, especially considering the relative lack of knowledge we still have about how our own intelligence works.
This makes a dangerous assumption, namely that our intelligence is the best. Mammalian brains are inefficient users of space compared to, say, avian or cephalopod brains. Assuming that the AI crowd is "overly optimistic" also makes the implicit assumption that the computer hardware the AI would live on will also never come far.
The human brain is a intricate network of hormones, neurotransmitters, and chemical electrical signals. And that's just describing the brain, all by itself, then their's the signals we receive from our senses(eyes, nose, touch, hearing, balance sense).
We also have a fair idea on how those senses work, and how our brains are wired up to recieve the data. This is not black-box magic.
There's more of course, there's all the varied parts of the brain, cerebellum, cerebrum, medulla, corpus collupsum,
The only parts of the brain that are related in any way to intelligence would be the cerebrum and the corpus collosum that connects its two halves. The medulla regulates autonomous bodily functions and takes no role in intelligence. The cerebellum is chiefly responsible for sensory integration and fine motor control.
then their's the way memory is stored,
Why should this matter for intelligence? Evolution drove memories to be encoded in the networks of neurons that also do work in cognition. The only reason for this is that's the most space-efficient way to do it, given the limitations imposed by the evolutionary baggage of having all the processing done on the surface of the brain. It doesn't matter how you store the information, so long as you can get to it when it's needed.
how various signals are altered depending on the various neurotransmitters in the brain.
Nice to know, yes, but in the end, it's all chemistry, and chemistry is an extension of physics, which can be simulated on a sufficiently powerful computer. Failing that, you can back up a few levels of abstraction and treat this messy chemistry as a black-box . . . i.e. we only care about the output as it relates to the input, without caring too much about the fine detail of the mechanisms of how we got there.
How certain signals are interpreted from where in the body, Fight or Flight responses, body brain interfaces(immune activity to a certain extent can be predilicated upon how a person feels).
Relevant for a medical simulation of the human body. Irrelevant for AI.
Then there's the body and how it affects the brain, then there's how the environment outside that affects the brain,
This is data. The environmental data we would care about, since we want our AI to interact with the outside world. The effects of the body don't matter so much.
then their's how our genes affect our brains, then their's and so on and so on.
Fine if you're building a human simulator. Irrelevant if you're building an AI.
I think AI researchers tend to think that just because we do it, a machine could do it, probably right, but for a machine to have true intelligence it would have to have a similar degree of complexity for it to have intelligence.
Our brains are messy and inefficient. Much of what goes on inside it is highly redundant. Strip away all that, and you're ultimately left with an expression x = f(n), where x is the output generated by the function f acting on the inputs n. We may not necessarily care about how exactly a neuron does it if we can encode a software or hardware function that can generate the expected values of x for the given values of n. And, again, you're assuming that we can never build something with a sufficient amount of processing power to do the job, even though nature managed to do so and did so generally by accident.
Its not just the brain, its the total system and the brain which makes us intelligent, you can't segment them off into parts and think that would acheive it.
Wrong. If I were to remove your head, your body would retain no measure of your intelligence. It would fall down, twitch a bit from random noise from what's left of your spinal cord, and then die, as the neural circuitry responsible for regulating it has been removed. If I were to pluck out your eyes and somehow implant a million electrodes in your visual cortex, and hook them up to a multi-megapixel digital camera, you'd not notice the difference. I could smash your inner-ear bones, and then wire you up with a cochlear implant and a microphone, and you would regain some measure of hearing. Intelligence resides soley within the brain.
Assuming of course it is even possible to have a AI intelligence(at least one that thinks with a computer speed), some evidence suggests that it is partly our slow electrical/chemical signals that allow our intelligence. In other words if a AI is developed it would think not that much faster then us.
Our intelligence arises from the complexity of our brains and their ability to process an enormous amount of information at any given moment. It achieves this through massive parallelization and massive interconnection between parallel elements. The only reason our brains are so slow is because the chemistry that drives them requires a very narrow range of temperatures to operate. We could drive up the speed at which everything in the brain occurs, and not affect our intelligence at all, except for the rate at which it occurs. However, we'd either die of starvation or suffocation very quickly, or we'd kill ourselves with our own waste heat. We're not limited to sloppy, slow organics. If we create an artificial intelligence of human or superhuman complexity, there's no reason to assume that it will move as slowly as we do, especially as we can employ more rigorous methods of boosting the processing rate of our computers, and the removal of the waste heat generated by their operation than is possible in a biological system.
*
