Michael Kanaan

You can apply the cutest algorithm you like to a non-random set of numbers and that will not make the resulting set random. If you feed in the same original set you will derive the same results. Furthermore, algorithms always commence with a rules-based approach and the same starting point will arrive at the same outcome however one might believe that some kind of spontaneous random mutation of the algorithm occurs during its processing. Lastly, not all problems are solvable by algorithmic means either because the problem is itself based upon a chaotic system in which the starting point of every possible variable can not be determined, or because the will fail to complete using a Turing machine.

Being able to manufacture a small network which appears to mimic the human brain, whilst a laudable achievement, does nothing to address the recent discoveries regarding the point at which a sentient being such as a human arrives at a conscious decision. That research suggests that, far from being merely a sophisticated computing engine, the human brain possesses a far greater degree of subtlety in respect of "free will" than any purported AI system could, at this stage, possible replicate.

Current AI is no more than patterns matching (Kanaan has a different view on it). Neural nets, rule based, other methods, it’s still pattern match. There is no motive for an AI machine, we pin our motives to the machine to get a result. We feel treat because pattern match is one of our brain best ability, but we also have others abilities like imagination (speculation restricted to narrow possible and desired outcome), social interaction, these just as examples. The fact that machine AI can’t have motive is some assurance machine won’t turn against humans on its own (which also begs the conclusion humans have the motive to go against humans, hardly news). The fact that AI is in essence pattern matching is bit troublesome for critical tasks because there can be errors (no rules can fit really perfectly). Humans have the most adaptable cerebral but feel treated by a pattern matching algorithm. Are we no better than a robot? Once “programmed” can’t change?

If one takes one of the important human senses, vision in my example, the brain does far more that merely "pattern-match". The processing involved does not just process the incoming information but supplies, I understand, over 50% of that information from its own internal resources including memory. This means that over a half of what you believe yourself to be seeing is generated through the processes of the visual cortex. This is several orders beyond "pattern-matching" no matter how sophisticated the underlying algorithm might be. Great works of art, music and groundbreaking discoveries in science are not the result of pattern-matching on flakes of paint, notes on a stave or mathematical symbols.… (meer)