Gigaom AI Minute – May 18

Transcript

One of the hopes with artificial intelligence is that it reaches a point where it can take on improving itself. This won't even require a general intelligence, just to the point that even a narrow AI can get better and better at what it does as it iterates.

Now, if it's getting better, then almost by definition, a self-improving AI will likely change its mind a whole lot. In this regard it's not going to be like us--we're stubborn, resistant to changing our minds even when the facts change. But an AI, improving itself, might think one thing one day and a new thing the next, and then back to the old one the next, and so forth. It might seem erratic to us as our ability to understand its ever more complex reasoning falls to the wayside.