Humans perform symbol manipulation using deep neural networks. Why would we expect that their artificial implementations (whether CNNs, SNNs, or some as yet unannounced algortihm) would not be able to do same?
Open letter to @ylecun: I have been explicit that I believe that symbol-manipulation is part of the solution to AGI; Hinton has ridiculed that idea. Where do you fit in? With me? W Hinton? If in between, where? The field would benefit from a clear statement of your view.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
-
-
-
We have no idea of the neural architecture at the microcitcuit level that people use for reading, formal logic (for those who could do it), etc; and IMHO no particular reason to believe that those circuits will be like the ones currently modelers use for imagine processing.
- 1 more reply
New conversation -
-
-
To me this is a "show - don't tell" exercise. this is why we do research/science and provide hard facts that go beyond speculation. nothing wrong with completely opposing opinions as it fuels research. In particular, I don't think this has to be a popularity contest.
-
I’m not taking a vote here, I am asking for explication. A large number of people have expressed interest in this debate, but on what I think is the most central point, there has been a lack of clarity.
- 2 more replies
New conversation -
-
-
Yes I strongly believe that AGI will need a multiplicity of approaches, incl. symbolic and ANNs. I also believe the disagreement highlights a major underlying problem: we have no general mathematical theory of intelligence. 1/2
-
Without a general theory of *natural* intelligence trying to build AGI is a bit like trying to do particle physics experiments with no standard model. Even w/o a theory AI researchers would benefit from deep study of natural/animal intelligence and evolutionary biology. 2/2
- 2 more replies
New conversation -
-
-
@tabithagold@g7 I don’t agree that ML has made symbolic AI obsolete. The two approaches need to be combined. - 1 more reply
New conversation -
-
-
Deep learning still requires a massive amount of hand coded engineering, human annotated data, and engineering data pipelines to do anything practical at scale - I think that's often overlooked. The support infrastructure around the model far exceeds the model code.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Both
#MachineLearning and#symbolic#AI are product of human intelligence. Even the most autonomous learning algorithms are able to do it because of "rules" build into it.#AI tech is a continuous not a battle between two factions. Let's work at ensure#responsibleAI#AIforGood.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.