so true @GaryMarcus whilst keeping the public illinformed about the outstanding scientific problems. The industry giants chase leaderboards that the media promote, which do not represent understanding of language.
this — digging into the details of how intelligence works — is exactly what mainstream machine learning has tried to sidestep of late, on the theory that general intelligence will “emerge” from big data and some sort of general learning algorithm. Thus far it hasn’t.https://twitter.com/ThePatHayes/status/1192934492509528065 …
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The process of 'discovering' an artificial version of natural general intelligence, in the same sense as artificial flight was 'discovered' by the Wright bros., has not started yet. They did not discover a flight simulator. They used an inorganic form of the natural physics.
- 8 more replies
New conversation -
-
-
In physics, we want TOE(theory of everything) which hasn't materialized. GI probably would follow the same fate
- 1 more reply
New conversation -
-
-
Expressed in our book The Humachine is that we cannot code our way into AGI even via brute force, neither are biological cognitive enhancement, whole brain emulation, or neural lace viable pathways. Instead, we should aim for organizational network ingelligence 1/2
-
That is, we can apply Kasparov's Law to leverage Moravec's Paradox so as to satisfy Nick Bostrom's definition of collective superintelligence. That is the only viable pathway on relevant time horizons.
- 1 more reply
New conversation -
-
-
Even if you managed to get emergent intelligence from a massive neural net or something, it still wouldn't answer the question of understanding intelligence.
-
true enough, but i can't bend my counterfactuals that far :)
- 1 more reply
New conversation -
-
-
It never will, in the sci-fi skynet sense. Modern AI is focused lately on clever ways of acquiring, organizing, and accessing data. A pile of books will never become self-aware or self-directed no matter how many books you put in the pile, nor how you arrange them.
-
In the long run an overarching motivational algorithm will need to be directly coded. It won't just "evolve" out of nothing. Having said that, I suspect that such an algorithm will be surprisingly simple to write. (At least, simple to write a simplistic one.)
End of conversation
New conversation -
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.