A major missing piece for AGI, is humans' ability to set their own goals and reward functions. E.g. feeling rewarded for winning a "game"https://twitter.com/fchollet/status/876227262437179392 …
-
In effect, we have the power to completely hack our own reward system, down to the ability to override e.g. pain avoidance
In that sense, "general intelligence" is not just the ability to find optimal solutions to externally-provided problems (making paperclips)
-
-
It is also the ability to figure out what goals you should pursue. Which in turn guides the development of your own intelligence
-
When thinking of humans in terms of RL, it's surprisingly hard to figure out where to draw the boundary between agent and environment.
End of conversation
New conversation -
-
-
Satinder Singh was discussing learning intrinsic rewards for RL back at ICML '09 http://web.eecs.umich.edu/~baveja/Papers/FinalNIPSIMRL.pdf …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.