Archive for posts tagged ‘ai’
I’m back to reading Structure and Interpretation of Computer Programs after a two-month break, and learning Lisp inevitably makes me daydream about being able to write artificially intelligent programs. Once I get better with Lisp, I’ll probably pick up a book by Peter Norvig and finally understand how hard it is, but for the moment I like to pretend that I can do it.
Anyways, my most recent thought is that even if you could have a program that was capable of making decisions that we normally assume only a human can do, would you trust it? I think that it could be trusted, but the deciding factor would be how long the computation took.
I haven’t looked into any psychology research on this (it’s my blog, I’m allowed to spew my thoughts without researching them), but I think humans would personify the program too much and expect difficult problems to take a long time to solve, regardless of how powerful the computer was. Consider Deep Thought in Hitchiker’s Guide to the Galaxy. That fictitious computer was far more powerful than anything we have today, yet it still took hundreds of years to determine the answer to life, the universe, and everything. Surely a program that determined if we should go to war would take at least a few months.
That’s why I don’t think that computers will ever be trusted to make decisions that are more than yes/no answers or engineering design problems that can eventually be tested. For instance, in the book Player Piano, a computer named EPICAC is in charge of determining what jobs people are qualified for along with other decisions that affect the economy. In real life, for a computer to be trusted with that kind of responsibility, the computation would have to run for as long as a human would normally take to solve the problem. If that’s the case, then what’s the point of using a computer at all?
Even if an artificially intelligent machine can prove that it can make difficult decisions in only a fraction of the time it takes a human, it will still be perceived as being “hasty.” Don’t get me wrong; there will be plenty of great uses of AI, but holding positions of responsibility probably won’t be one of them.