The biggest problem in AI? Machines have no common sense. Great video by @GaryMarcus.
#ArtificialIntelligence #MachineLearninghttps://bigthink.com/videos/how-smart-are-robots-ai …
-
I agree that space, time and causality are different domains from the game of Go. However, fundamentally, our intuitive understanding of space, time and causality are due to the amortized inference that evolution has developed for us over the eons. That's the breakthrough of DL.
this is totally confused. major emphasis in deep learning has been in trying to learn things bottom up from data, whereas nativists stress power of what evolution has done, to build strong starting points - that are often in the deep learning community considered to be "cheating"
-
-
It is normal practice in DL research to begin with the least amount of assumptions in the design of their networks. We of course know that this is not how biological brains work. Brains already have representations that are fine tuned over the eons. This is amortized inference.
-
The real problem is how to get from one amortized inference to more advanced amortized inferences. How do you get a pretrained network to extrapolate into new domains. I call this "Artificial Ingenuity" https://medium.com/intuitionmachine/the-key-to-general-intelligence-is-the-mechanism-of-artificial-ingenuity-dad26bc99043 … . Where I do discuss DL limits.
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.