The Path To AGI
AI skeptic: “but it doesn’t do X!”
AI researchers: “good point, let’s figure out how to do X”
3 months later..
AI skeptic: “but it doesn’t do Y!”
AI researchers: “good point, let’s figure out how to..”
keep it up! keep the ideas flowing!
AI skeptic: “but it doesn’t do X!”
AI researchers: “good point, let’s figure out how to do X”
3 months later..
AI skeptic: “but it doesn’t do Y!”
AI researchers: “good point, let’s figure out how to..”
keep it up! keep the ideas flowing!
Comments
This divide will keep going, because we don’t know if the number of X tasks to accomplish is feasible / attainable
It's a useful objective so that people get more objective functions to optimize towards (a series of specialized tools fulfilling certain purposes), but never one *thing* that happens.
Is it useful in this domain? Great. Is it not useful in this other one? Well then, how do we make our tools useful for it?
These specialized tools, will eventually either be useful or not useful in an engineering sense.
I think it's hard to judge in the speculation phase (which we are in now) what happens during the dispersion phase. i.e., the microchip innovators could have never forseen Alexas.