Is Reinforcement Learning the possible solution for General Artificial Intelligence or is there something we have not figured out yet?
-
-
-
Concept formation.
-
Concept Formation as in?
-
@demishassabis Roughly the process that allows humans to go from perceptual data (visual, audio, etc), and form words (concepts) that can be used to think and talk. Basically what appears to be the key cognitive difference between humans and other animals.
End of conversation
New conversation -
-
-
Demis -- when are you going to ditch those Google guys and get back to the job at hand building the first GAI? I've had my bet on Deep Mind for years --- you guys were on fire until you got swallowed up by The Borg.
-
Google is extremely politically biased and has now coined themselves as "The Good Censor". An organization that feels they can censor for the "greater good" should NOT be in control of an AGI.
-
Google suffers from a "not built here" superiority complex -- brutal for a super talented UK team.
End of conversation
New conversation -
-
-
This is at the Rothschild's foundation. Once AGI is created, it will be used to shape society to benefit the elite. It will be at all social junctions to control the flow of information. This AGI would be their thought-police. Rewards for the elite are too high to not do this.
-
I wouldn't worry. Hassabis just pontificates. He's clueless about how true intelligence works. His programs play fancy computer games but he can't build a robot that can walk into a kitchen and make a sandwich.
@DeepMindAI has as much chance of figuring out AGI as my dog. -
Nobody knows how to create AGI, it's still something to be created. The trait an AI researcher needs to discover the missing ingredients is open-minded essential, he seems to have this. If you watch his video he demonstrates he is willing to accept new ideas that lead to AGI.
-
No he is not. He is married to deep learning and backpropagation. I have read several of his papers.
-
BTW, Hassabis is someone who pretends to know a thing or two about neuroscience but he does not seem to know that learning in the brain (Hebbian learning) is based on timing. He has pushed the nonsense that biological learning is a form of reinforcement-driven deep learning.
-
As a Christian, I am hostile to AGI researchers in Silicon Valley and academia. I think they are a bunch of clueless, elitist jackasses wasting money. I have argued that intelligence cannot be based on representations because the brain is not big enough. They're not listening.
-
Moreover, and I've said this many times, the brain can instantly see a new complex object or pattern that it has never seen before, i.e., with no prior representations. Deep learning is the learning of representations. Draw your own conclusion about its relevance to intelligence.
End of conversation
New conversation -
-
-
I'm creative as fuck.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Facinating lecture...Hearing the commentators talk about Move 37 was reminicent of how sporting legends become known for their signature moves. Can you see the day when AI's will develop their own unique styles?
- 1 more reply
New conversation -
-
-
Yes, Machine Psychology (communication [NLP], conceptual reasoning, commonsense, imagination) seems like the final frontier. I think we might be really closing in on AGI now Demis! (At least in the sense of having an 'in-principle big picture' of how it all might work).
-
We start with 'Computational Logic', which I think involves two levels of knowledge representation: *Many-Valued Modal Logic (Low level) *Formal Language Theory (High-Level) Wiki-book: https://en.wikipedia.org/wiki/User:Zarzuelazen/Books/Reality_Theory:_Computational%26Non-Classical_Logic …
-
And I think we need to extend computational logic (which is about knowledge representation) to deal with active imagination and reflection (harnessing knowledge towards internal goals). The result I think is something like a 'Conceptual Tree'.
-
Computational Logic >>>> Machine Psychology Two levels for machine psychology as well, I suggest: Ontology&Language Models (Low Level) Scientific Theorizing (High Level) Linguistics parse trees and argument trees? (low and high levels) Wiki-book: https://en.wikipedia.org/wiki/User:Zarzuelazen/Books/Reality_Theory:_Machine_Psychology%26NLP …
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.