I think deep learning researchers agree about constructing interpretations, but we want those interpretations to be part of a differentiable architecture rather than an explicit symbolic-sub-symbolic hybrid. Time will tell which approach is more effective.
What @tdietterich says here echoes central argument of http://rebooting.ai : “deeper understanding requires building an interpretation of the query/story/article”
#deeplearning has not come close to this (as my 2015 quote anticipated)
And its why we MUST consider hybrids.https://twitter.com/tdietterich/status/1187609058376704000 …
-
-
- 1 more reply
New conversation -
-
-
I'd go further than that. *Learning* is the integration of knowledge into a mind. So in ML, not to mention DL, the "L" is only a
#WishfulMnemonicThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I disagree with the hybrid approach. I'm a few months away from a critical stage in my project where I can demonstrate ML/DL isnt part of the AGI solution. My AGI will be able to create general representations of anything and use them to classify any novel object. With ZERO ML.
- 1 more reply
New conversation -
-
-
No, we must throw away both DL and GOFAI. They fail to understand, not just language, but everything. Pattern classification is not understanding. It's just a rule: IF pattern A THEN label X. Symbols don't help. They fail because there's no universal building block of knowledge.
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.