Gigaom brings you our unique analysis and commentary on the present and future of AI.
The kind of AI that we have today that we're excited about is machine learning. That's where we take a bunch of data about the past, we study it, and we try to make predictions about the future. Now, are there places where that technique doesn't work? It turns out, actually, there are a number of them. In the end, though, is machine learning fundamentally flawed for the most interesting kinds of problems? Isn't it the case, perhaps, that machine learning cannot be creative--that machine learning cannot create art?
It's no doubt that a machine learning program can take a photo and Van Gogh-ize it, or it can turn it into an impressionist painting. It can take a small piece of music and Beethoven-ize it. So it can do things that are completely derivative of something else, but can an artificial intelligence make something completely new? It may be able to in the future, but right now the basic idea behind machine learning is that if you give it enough examples of the past it can make reasonable predictions about the future. That is almost the antithesis of how creativity works.