Twitter | Pretraživanje | |
Dileep George
Founder at Vicarious AI ( ). Previously Founder and CTO at Numenta. Triply EE (BTech IIT-Mumbai, MS&PhD Stanford)
1.187
Tweetovi
359
Pratim
2.184
Osobe koje vas prate
Tweetovi
Dileep George 5 h
Odgovor korisniku/ci @WiringTheBrain
Forget the amoeba, human free will is still being debated :)
Reply Retweet Označi sa "sviđa mi se"
Dileep George 5 h
Odgovor korisniku/ci @WiringTheBrain
to be considered agency, it should be able to stop doing that at will when the yeast cell is present. Can it? Or is it a hardwired stimulus-response mapping that we are anthropomorphizing?
Reply Retweet Označi sa "sviđa mi se"
Dileep George proslijedio/la je tweet
Brad Wyble 3. velj
Insanely weird bistable photo. Is this a dog or a weird looking goat? (not sure where the photo is from, can't provide credit)
Reply Retweet Označi sa "sviđa mi se"
Dileep George 3. velj
Odgovor korisniku/ci @smjain
Thank you!
Reply Retweet Označi sa "sviđa mi se"
Dileep George 2. velj
Odgovor korisniku/ci @rplevy @MelMitchell1
no, it is just an accurate depiction of my drawing disability.. :)
Reply Retweet Označi sa "sviđa mi se"
Dileep George 2. velj
Odgovor korisniku/ci @iAhmedMaher
thanks for making it!
Reply Retweet Označi sa "sviđa mi se"
Dileep George proslijedio/la je tweet
Ahmed Maher 2. velj
If u ever wondered about the inner workings of 's RCNs, I just published an article on it:
Reply Retweet Označi sa "sviđa mi se"
Dileep George 2. velj
This blog on RCN model is just excellent with amazing visuals! I highly recommend it, especially if you are interested in generative visual models based on the visual cortex.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 2. velj
Odgovor korisniku/ci @MelMitchell1
take on the trolley problem.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 30. sij
Odgovor korisniku/ci @summerfieldlab
Really nice! For your module on generative models, I would like to suggest our RCN paper, which has exactly this kind of shape-object factorization and is an example of a neuroscience inspired predictive-coding model that works in practice.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 29. sij
Odgovor korisniku/ci @Abel_TorresM @GaryMarcus
yes, I think it is an interesting problem to work on. We dabbled on that problem some time ago, but haven't really pursued it in a while....
Reply Retweet Označi sa "sviđa mi se"
Dileep George 29. sij
Odgovor korisniku/ci @Abel_TorresM @GaryMarcus
reachability analysis a static setting is easy via message-passing. Just propagate forward from the current state until you hit the target or exceed the graph diameter. Sokoban settings can be arbitrarily hard because you can move blocks out of the way etc.
Reply Retweet Označi sa "sviđa mi se"
Dileep George proslijedio/la je tweet
Dileep George 29. sij
Odgovor korisniku/ci @GaryMarcus
Nice video. This is exactly the innateness that is used in our paper "Different clones for different contexts" () and in our cognitive maps paper that learns spatial layout from from ordinal sequential info
Reply Retweet Označi sa "sviđa mi se"
Dileep George 29. sij
Odgovor korisniku/ci @GaryMarcus
Nice video. This is exactly the innateness that is used in our paper "Different clones for different contexts" () and in our cognitive maps paper that learns spatial layout from from ordinal sequential info
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @roydanroy @pacificlion123 i 2 ostali
Amelia Bedelia?
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @FelixHill84 @egrefen
also message-passing for inspired by the interactive activation model fwiw
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @FelixHill84 @egrefen
We do it in our paper using message-passing. It is approximate, and there are no theoretical guarantees, but works really well in the settings we have tried. Also it is very fast.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @FelixHill84 @egrefen
it buys an a mechanism to do "dynamic inference" in an economic way. By dynamic I mean it is not amortized inference that requires exposure to all the different compositions during training.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @FelixHill84
"reality of the letters", doesn't really exist :). I can show you lots of characters that are created purely via subjective contours.
Reply Retweet Označi sa "sviđa mi se"
Dileep George 28. sij
Odgovor korisniku/ci @FelixHill84
thanks! :). I think I understand what you mean. Resolving all that context might be equivalent to abductive inference though, right?
Reply Retweet Označi sa "sviđa mi se"