The 'Trace' section of the The Matrix Cookbook translated into tensor diagrams.

The nice thing about the diagrams is that they show why the formulas are true.

thonking.ai/p/strangely-ma… 🤯

h/t Justin Pinkney

Goosebumps.

I've been working on a new product: MakeDistribution.com!

It's the best solution to creating a subjective probability distribution, i.e. one that reflects human judgement rather than being fit to data.

Sounds simple, but most existing solutions have important flaws (1/15)

a new blog post, because it is Saturday.

<Fixing DPO but I have a dinner reservation …> kyunghyuncho.me/a-proper-prefe…

ShengShu Technology with Tsinghua University, officially released China's first long duration, high consistency, and high dynamics video model 【Vidu】.
Original Diffusion and Transformer fusion architecture U- ViT
16s with 1080P

Has anyone written a blog post yet comparing the reapers in mass effect with training models on all possible human data? 🤔😅

Here is my full statement on Google's response to this week's newsletter. They have failed to refute any statement that I made, and in doing so helped me find yet more emails from the DOJ that show Google's ads team made search worse to make money.
wheresyoured.at/in-response-to…

After a longer and more careful read of
blog.normalcomputing.ai/posts/2023-09-…

I wonder if this is what RAG 2.0 looks like. Went all the way back to 'Memory Models' to get the full picture.

News are erupting in Germany that the parliament's decision to go through with the nuclear phase-out was partly based on false information that seems to have been deliberately falsified.

cicero.de/innenpolitik/r…

Come work with us for 6 months on synthetic data research! Reach out to amazing Cate if you are interested 🔥

In the last 24 hours, two of our leading media sources (NYT & WSJ) have produced research pieces with hard data on why the U.S. should ban TikTok.

Here is the NYT's killer chart. Tiktok systematically censors 80-99% of posts on certain viewpoints, relative to Instagram.
1/3

Very happy that one of my two ICLR 2024 blog posts has been highlighted as a spotlight (blog post) 🥳🥳🥳

It's 'Bridging the Data Processing Inequality and Function-Space Variational Inference'.

I'm quite proud of it, so let me give you a quick tour:

Arrow supports two file formats: the 'file' format and the 'streaming' file format which are not the same. But which file extension is meant for which file format? 😅🙈

Acc to stackoverflow.com/questions/6791… it seems that .feather matches the arrow 'file' file format while .arrow is the

The fact that AI can generate a realistic 1950s Super Mario movie with just a few prompts is mind blowing. We are so not prepared for what AI can do. 📎👀

AI Notkilleveryoneism Memes ⏸️ , Elon Musk

A bit of Oxford uni drama. I’m looking for sources to verify or contradict my arguments and data.

Someone posted a call to join a protest/event in support of the “Gaza solidarity encampments in the US” in the Exeter MCR WhatsApp group, and people got upset when I expressed hope

Data is all we need! 👑 Not only since Llama 3 have we known that data is all we need. Excited to share 🍷 FineWeb, a 15T token open-source dataset! Fineweb is a deduplicated English web dataset derived from CommonCrawl created at Hugging Face! 🌐

TL;DR:
🌐 15T tokens of cleaned