for computer scientists, existence means to be created. before GH sees a hybrid model, he will not believe it
Geoff Hinton dismissing symbol-manipulation and ignoring possibility of hybrid models, at G7 meeting earlier this week. When I referred to ”deep learning imperialism” it is exactly this, and the conflation of Google Translate with real language understanding, that I had in mind.https://twitter.com/tabithagold/status/1070736319901519876 …
1:57
Drop the definitive mic
Geoffery Hinton explains the difference between symbolic AI and deep learning to great applause from the @g7 audience. A must listen: pic.twitter.com/HbeQZX9QkH
Geoffery Hinton explains the difference between symbolic AI and deep learning to great applause from the
Show this thread
-
-
-
alphago is a hybrid model
- 3 more replies
New conversation -
-
-
Interesting in light of--"If we could embed symbol reps & structure manipulating operations within a distributed, subsymbolic architecture, then very powerful .. fault tolerant high-level reasoning/planning systems could be created" (Touretzky & Hinton, 1988)
-
yes a great paper from an earlier, more open-minded Hinton
End of conversation
New conversation -
-
-
I believe any AI research deserves funding. The key here is what percentage is Europe spending on symbolic vs connectionist. Does anyone have an idea?
-
How about in the world as a whole, industry included? 80% deep learning? 90%? How much on hybrid models? 3%?
- 1 more reply
New conversation -
-
-
More interesting is what he hints at toward the end. It would be somewhat amusing if AI approach turns out to be cultural with the Europeans taking a more rationalist (civil law) view of AI vs the British/North American empiricist (case law) view.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Initially, I was a sceptic of symbolic manipulation but now strongly believe syncretism or hybrid approaches are the only way to realise the full potential of AI, at least in medicine:https://www.linkedin.com/pulse/moving-from-probability-inference-causal-hybrid-approaches-reddy/?published=t …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Sad and dangerous for science as a whole. And for AI in general. Never expected to hear this from him. Tons of counter examples. Tons of evidences showing that both approaches have pros and cons.
#dontblindlyjumpintotheMLbubbleThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The idea that Google Translate models linguistic competence is just reanimated behaviorism in its most brute form. Unrelated to realist questions concerning the generative mechanisms enabling a child to generate passives or polar interrogatives.
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.