Introducing Nested Learning: A new ML paradigm for continual learning that views models as nested optimization problems to enhance long context processing. Our proof-of-concept model, Hope, shows improved performance in language modeling. Learn more: goo.gle/47LJrzI @GoogleAI

AI and the Future of Work piped.video/McYBgZrORi4?si=bR4D… via @YouTube

a lot more fake LLM powered accounts are popping up and these are a little bit sophisticated. unfortunate

MTG just torched her own party: “Health care crisis? Ignored. Wages? Flat. Bills? Sky high. And you think this wins midterms?” She says her electricity bill is up $100 since last year — and accuses Republicans of abandoning “America First.” When Marjorie Taylor Greene starts sounding like Bernie Sanders, it’s not bipartisanship — it’s a five-alarm warning that the base is fed up.

This is one of the most disgusting and vile things I have ever read, and that's saying a lot as a woman growing up in chess. "Where were all of you when Danya was alive and unwell?". You are the CEO of the governing body of chess and in your first real comments since the passing of one of our greatest talents, coaches, and ambassadors you decide to blame the public when you have absolutely no idea who Danya had been in contact with? Where the f*ck were you? Certainly not protecting your players from harassment because I'm sure that might have interfered with your protection of FIDE President Dvorkovich, close friends with Kramnik. You both accuse his friends knowing NOTHING about the support he received and remove any responsibility from yourself at the same time. Truly pathetic. We aren't oversimplifying anything. We know the damage Kramnik did to Danya because he told us in some of his final painful words. Trivializing the world's call for justice against Kramnik's malicious campaign of harassment isn't "virtue signaling", but I'm not at all surprised this would be your take given your morals have been bought out for many years now. But you are right about one thing. We absolutely didn't do enough while Danya was with us, we were not vocal enough about the failures of FIDE to uphold it's own bylaws, and that's what makes me the most sick. It's a mistake I won't make again. You are clearly unfit to lead FIDE. You are a disgrace to Chess and everything it should stand for. Resign immediately.

“Bed is stuck in inclined position because of AWS outage” I really thought this must have been a joke

Using custom-trained LLMs and > 1k 4090s to visualize 100k scientific research papers in latent space 🌐 DM me for early access 🔜

🚨 New paper out! “FineVision: Open Data Is All You Need” 🥳 We unified 200+ data sources into 24M samples. That’s 17.3M images and 9.5B answer tokens, the largest open VLM dataset ever released. All fully documented, reproducible, and available for everyone. And there's more! 🎢

I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter. The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language person) is whether pixels are better inputs to LLMs than text. Whether text tokens are wasteful and just terrible, at the input. Maybe it makes more sense that all inputs to LLMs should only ever be images. Even if you happen to have pure text input, maybe you'd prefer to render it and then feed that in: - more information compression (see paper) => shorter context windows, more efficiency - significantly more general information stream => not just text, but e.g. bold text, colored text, arbitrary images. - input can now be processed with bidirectional attention easily and as default, not autoregressive attention - a lot more powerful. - delete the tokenizer (at the input)!! I already ranted about how much I dislike the tokenizer. Tokenizers are ugly, separate, not end-to-end stage. It "imports" all the ugliness of Unicode, byte encodings, it inherits a lot of historical baggage, security/jailbreak risk (e.g. continuation bytes). It makes two characters that look identical to the eye look as two completely different tokens internally in the network. A smiling emoji looks like a weird token, not an... actual smiling face, pixels and all, and all the transfer learning that brings along. The tokenizer must go. OCR is just one of many useful vision -> text tasks. And text -> text tasks can be made to be vision ->text tasks. Not vice versa. So many the User message is images, but the decoder (the Assistant response) remains text. It's a lot less obvious how to output pixels realistically... or if you'd want to. Now I have to also fight the urge to side quest an image-input-only version of nanochat...

I doubt I’ll run for POTUS, but I appreciate the support @jack. I’d be happy if we could just get 4 or 5 more voices in Congress who don’t always just do what their party tells them.

arxiv.org/pdf/2510.01395 Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence If you remember anything, remember this. If you value your own intelligence, interact with AI systematically. Until alignment and verifiers get better, you will dig your own rabbit hole

python, pithon

The surest way to screw up the world’s best technical school is to let feds tell them how to run it. Congrats to my alma mater for turning down a bribe to let the executive branch dictate what happens on its campus. A lot of things are wrong in 🇺🇸, but MIT is not one of them.

“Massie Introduces Bill to Stop the Government From Propagandizing Americans” thenewamerican.com/features/…

Contrary to what he says, @SpeakerJohnson is doing everything he can, including delaying the swearing in of the most recently elected member of Congress and spreading misinformation about the legislation, to block a vote in Congress on legislation to release the Epstein files.

love it that i can finally browse a pretraining corpus and everything is somewhat interesting.

Nearly the same prompt, imaginary science-fiction movie in the style of Visconti, Stable diffusion August 2022 vs. Aurora January 2025.

The government is in full shutdown and the Republicans are refusing to call the House back into session. Want to know why? Because we have secured the final vote on releasing the Epstein Files and they don’t want it out. Call GOP and tell them to swear in @AdelitaForAZ.

stop watching fox news if you care about this country.

🚨 DEMS Launch INVESTIGATION into Trump $17 TRILLION THEFT?!!! piped.video/2qqd98OjYnk?si=4_gq… via @YouTube