iiinova — where your digital self keeps evolving

Dileep George @dileeplearning @dileeplearning.bsky.social · 15/05/2025

somehow chatGPT understand my opinion about successor representations? 4/

Dileep George @dileeplearning @dileeplearning.bsky.social · 15/05/2025

I didn't mention partial observability specifically, so it is impressive that this was picked up. Looks like we did something right in our CSCG paper in making this explicit? 3/

100

Dileep George @dileeplearning @dileeplearning.bsky.social · 15/05/2025

It is quite impressive that chatGPT picked up these nuances, picks up a relevant quote from the paper and even emphasizes portions of the response. 2/

100

Dileep George @dileeplearning @dileeplearning.bsky.social · 15/05/2025

This paper turned up on a feed, I was intrigued by it and started reading... ..but then I was quite baffled because our CSCG work seem to have tackled many of these problems in a more general setting and it's not even mentioned! So I asked ChatGPT... ...I'm impressed by the answer1. 1/🧵

1111

Dileep George @dileeplearning @dileeplearning.bsky.social · 29/04/2025

Wow, very cool to see this work from Alla Karpova's lab. She had shown me the results when I visited @hhmijanelia.bsky.social and I was blown away. www.biorxiv.org/content/10.1... 1/

1335

Dileep George @dileeplearning @dileeplearning.bsky.social · 27/04/2025

It's kinda obvious. #AGIComics has already figured out which brain region is the most important. 😇

3245

Dileep George @dileeplearning @dileeplearning.bsky.social · 26/04/2025

And whether top-down't influence is multiplicative or not is very context-dependent. (this is also what is seen in neurobiology). www.science.org/doi/10.1126/...

150

Dileep George @dileeplearning @dileeplearning.bsky.social · 18/03/2025

.... "anxiety", "emotional states" etc. are in quotes, and then used freely. "stupid" article, in my "humble" opinion.

160

Dileep George @dileeplearning @dileeplearning.bsky.social · 18/03/2025

This research is "nonsense". The reviewers and editors that let this pass should be "ashamed".

2130

Dileep George @dileeplearning @dileeplearning.bsky.social · 14/03/2025

It was fun to visit Weinan's lab @sunw37.bsky.social at Cornell to give this talk. Exciting to see the experiments he's cooking up, and we had lots of fun discussions about PFC, hippocampus, consciousness and meditation! Part of the talk content is here: blog.dileeplearning.com/p/space-is-a...

2182

Dileep George @dileeplearning @dileeplearning.bsky.social · 07/03/2025

Not True for the papers above. The map is not being created by human in the loop and the organism itself doesn't "read the map". And this is not just our work, many other recent papers on cognitive maps have similar properties too.

130

Dileep George @dileeplearning @dileeplearning.bsky.social · 07/03/2025

yes. And the papers above do offer this. 1) "maps" are induced as latent graphs from sequence of sensory aliased sensory observations. 2) They perform the following computations (non-exhaustive): - inferring the current latent state based on history. - planning computations for reaching goals.

120

Dileep George @dileeplearning @dileeplearning.bsky.social · 07/03/2025

....yet to be found? I don't think so. At least the papers above provide exactly these mechanisms.

120

Dileep George @dileeplearning @dileeplearning.bsky.social · 10/02/2025

Wait ….deriving “functional networks” is not the same as mechanistic wiring.

020

Dileep George @dileeplearning @dileeplearning.bsky.social · 10/02/2025

enormous interconnection is not required to have confounders ... enormous interconnection is a correlation you are latching on to :-)

120

Dileep George @dileeplearning @dileeplearning.bsky.social · 07/02/2025

Fun fact: I sat on this paper for months before submitting because I was terrified of dealing with the review process. But then the reviews surprised me in a positive way.... I like this comment particularly because I drew most of the figures! :-)

2190

Dileep George @dileeplearning @dileeplearning.bsky.social · 23/11/2024

@davidchalmers.bsky.social is on bluesky. Follow him and tell him not to leave a vacuum in philosophical thought #AGIComics

2247

Dileep George @dileeplearning @dileeplearning.bsky.social · 21/11/2024

Amidst all the arguments, its important to remember science is all about love … #AGIComics

4191

Dileep George @dileeplearning @dileeplearning.bsky.social · 20/11/2024

heck, I can even give you a spiking neuron implementation if you want.

000

Dileep George @dileeplearning @dileeplearning.bsky.social · 20/11/2024

this is a picture from Jay McLellan's paper......connectionist or not? Maybe not according to the new standards...

100

Dileep George @dileeplearning @dileeplearning.bsky.social · 19/11/2024

For example, looks me what is meant here is fitting the data.

000

Dileep George @dileeplearning @dileeplearning.bsky.social · 19/11/2024

So here is a challenge: We know that visual cortex is NOT convolutional, yet convolutional models currently offer the best fit for the data. What is the neuro-connectionist explanation for this? In the iteration mentioned in the paper, when will 'non-convolutional' representations win?

161

Dileep George @dileeplearning @dileeplearning.bsky.social · 19/11/2024

According to @tyrellturing.bsky.social I was mischaracterizing the paper by the figure, but reading it, I think am actually characterizing it accurately.

490

Dileep George @dileeplearning @dileeplearning.bsky.social · 19/11/2024

20 years ago I was at the first ever Cosyne meeting as a grad student and I had a binder with me that was the first draft of Jeff’s book. I remember emailing him from Cosyne that the book will inspire a lot of folks in EE/CS folks to study brains…

3271

Dileep George @dileeplearning @dileeplearning.bsky.social · 18/09/2024

It's not just you who's confused about successor representations (SR), advanced chatbots are too, and I definitely don't blame them. They just need to read my blog: blog.dileeplearning.com/p/a-critique...