HN Explorer

Ask HN: Is General purpose AI going to sneak up on us?

5 points by elisharobinson ↗ HN
I have been following the ai space for a few years now and everyday there seems to be a breakthrough in fields people believed impossible to solve by AI systems. Although the general consensus is that good AI is quite a bit far away ,could it be we are vastly underestimating the capabilities of self learning systems. from my view all the basic ingredients for some form of AGI already exist.So i fear that we might create AGI as in spontaneous combustion of gas in a hot day.

2 comments

[ 2.8 ms ] story [ 16.2 ms ] thread
alistproducer2
Nope. There's no single algorithm that will lead to General intelligence. As long as the apex of our ML techniques is parameter optimization via gradient decent, there will be no critical mass point. Currently, it's possible for someone to take a modular approach to general intelligence. And someone may very well be working on that and will spring such a system on the world but because each subsystem is only as good as the state of the art, and if the systems were that good the developers would try to capitalize on such a system immediately rather than waiting for all the other subsystems to surpass the state of the art, we would have some warning that the field was advancing; hence no surprise general intelligence.
mindcrime
It's pretty hard to say if it will or not, but I don't think there's any question that it could. What's even harder to guess is when that might happen if it did.

All of that said, there are enough people working on AGI on a day-in / day-out basis who would probably recognize the incipient stages of a real AGI, that I don't worry a tremendous amount about being surprised by the emergence of AGI. But on the flip-side, some "kid in a garage" operating outside of the mainstream of industry and academia could be the one who makes The Big Breakthrough, so there's no way to be totally sure...