This is also from GPT-2. But I'm not sure either of us have demonstrated anything important about machine cognition.pic.twitter.com/yDK081Hg5a
icymi, this is the core problem with current ML approaches to natural language understanding, in a single tweet with two images:https://twitter.com/GaryMarcus/status/1194081788307496960 …
-
-
-
do me favor, try ten more times, and please tell me the % correct. (if you want, do the same with
@Wolfram_Alpha) - 2 more replies
New conversation -
-
-
Is computational heuristics part of understanding though? Even if it was you still need to "interpret" data correctly, which in itself involves "understanding" per se. It is likely though that hybrid approach (predictions plus heuristics) could get us further than we are todaypic.twitter.com/JkZr817Z48
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
To be fair, 16 bags of food for 5 boxes is a really sweet deal.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Turing said he didn't need to UNDERSTAND German to predict/translate its sequences. Now, if prediction isn't understanding & intelligence ... someone clever will need to OVERWRITE 2300+ YEARS of Aristotle-Leibniz & 250+ years of Bayes-Boole.
@ylecun@geoffreyhinton@navalpic.twitter.com/LiR7f5vsCm
-
Those missing mechanisms can be invented from ancient Chinese. It CANNOT be derived from Aristotle's logic through Markov-Viterbi-VonNeumann-Weiner-Chomsky-McCarthy-Vapnik et al. So SV is "up s*** creek without a paddle".
@albertwenger@argodesigned@kaifulee@BaiduResearchpic.twitter.com/YglGjOYIeJ
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.