Gigaom AI Minute – October 7

Transcript

Assuming for a moment that an AGI is possible, how might we build one? There is no agreed-upon answer to that question. I find that when I talk to people in the industry, they are split 50/50 on whether our present techniques that we're using to build out narrow AI will eventually result in a general intelligence, or if an AGI is fundamentally completely different than what we're building. This uncertainty makes AGI different than, say, putting a human on Mars. We know broadly how to approach it and while there are all kinds of questions that would need to be sorted through--like ship size, crew, duration, propulsion--we know how to get to Mars.

AGI is entirely different. No one knows how to build an AGI, or, more precisely, there is no agreement on it. This should not be too surprising. We don't know how "real" intelligence works either. We don't know how our brains, let alone our minds, work, and even if we understood those things with clarity, we don't know how we would map that knowledge to computers, or even if that is the right approach.