Amazing that Jürgen Schmidhuber gave this talk back in 2012, months before AlexNet paper was published. In 2012, many things he discussed, people just considered to be funny and a joke, but the same talk now would be considered at the center of AI debate and controversy. Full talk:

Slides for my lecture “LLM Reasoning” at Stanford CS 25: dennyzhou.github.io/LLM-Reasoning-… Key points: 1. Reasoning in LLMs simply means generating a sequence of intermediate tokens before producing the final answer. Whether this resembles human reasoning is irrelevant. The crucial

The best logical explanation of entropy. (It only took me 11 months to find an hour to read it). arxiv.org/abs/2409.09232

Very proud of our team. This feature deploys a model that is both the largest image-2-image model we've ever put in Pixel; and also the first diffusion model we’ve ever run inside the Pixel Camera.

Mutation Series 5 | Mutation 12/26, 2014 by Mario Klingemann💧💦

x.com/i/article/1961…

Now's a good time to be a problem.

Over the past year, my lab has been working on fleshing out theory/applications of the Platonic Representation Hypothesis. Today I want to share two new works on this topic: Eliciting higher alignment: arxiv.org/abs/2510.02425 Unpaired rep learning: arxiv.org/abs/2510.08492 1/9

Gary Marcus Gary’s right — “distribution shift” is one of AI’s biggest unsolved problems. In simple terms, it means this: when an AI system is trained on one type of data but then encounters something even slightly different in the real world, its performance often collapses. It’s like

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

Peter Voss 🤷‍♂️ while having to deal with size, occlusions, and photometric variations that have nothing to do with the meaning.

Dileep George I know it’s popular to hate tokenizers, but visual representations (which are also tokenized) bring a lot of messiness as well. Aspect ratios, cropping, resolution, brightness, etc. Sure, models learn to deal with that but it requires lots of data to make them robust wrt these.

Q2: Which languages actually help each other during training? And how much? 🌟Answer: We measure this empirically. We built a 38×38 transfer matrix, or 1,444 language pairs—the largest such resource to date. We highlight the top 5 most beneficial source languages for each

A new episode of The Information Bottleneck podcast!🎙️ This week we talked with Randall Balestriero (Randall Balestriero), assistant professor at Brown University, about Joint Embedding Predictive Architectures (JEPA) 🥳🥳🥳

Congrats to Google on Gemini 3! Looks like a great model.

Wouldn't it be great if neural networks knew a spiral... as a spiral? 🍥

YEEZY set the trend yeezy.com

Elon Musk Grok Elon, you once posted that Jürgen Schmidhuber invented everything. New Grok imagine, as a truthseeking AI, seems to agree that a lot of people are standing on the shoulders of giants. Credit where credit is due.

Earth's rotation in space by stabilizing the Milky Way galaxy. 📹 Aryeh Nirenberg

Building fully native Android apps is so easy! 1. Install Android Studio 2. Open Android Studio and start a blank project 3. Install Google Antigravity 4. Point Google Antigravity at the folder created by Android Studio 5. Set the agent to Gemini 3 Flash and tell Google Antigravity what you