Al Fin has had several posts about artificial intelligence on his blog over the past one or two weeks regarding the limitations, and in his opinion wrong directions taken in programming an ai equal to the human brain.
Personally, I think trying to program ai is a stupid idea. I was going to use a five dollar phrase like "patently bad", but stupid works just fine.
It is mental masturbation in it's highest form. I'm not saying that some highly useful algorithms won't fall out of it, but I think the unintended consequences could be horrific.
To write an algorithm that mimics the logical functions of the brain is doable I think.
Imagine a highly logical ai with no compassion, no empathy, no judgement, and no love.
Imagine that AI chauffeuring you in your car. It is impossible to write code to cover every conceivable and inconceivable circumstance.
I felt compelled to post a comment at Al's place regarding this, but couldn't bring the words together so I let it go. But apparently my brain did not.
I was reading blogs and happened to look up at the TV this evening, where I saw a robot limping - the terminator movie..a robot limping...
This brought my random thoughts together regarding AI.
And I thought, g'head, teach a robot to limp in spite of an injury/malfunction.
Or how about teach a computer to solve a problem with insufficient information and tools. Teach an ai how to wake up in the middle of the night with the answer to the problem.
I'm not saying we shouldn't try, but my gut tells me it is a fools errand. Either from the standpoint that it is unattainable, or unintended consequences.
postcript: I had to edit AI as the font made it appear that Al was chauffeuring your car, due to the font
Edge cases are a bitch. You could argue that everything interesting in AI is an edge case.
ReplyDelete