Explain the Normal Distribution to a Gym Bro.👇 The normal distribution is a continuous probability distribution that plays a central role in probability theory and statistics.

Reading the H3 paper this afternoon! Interviewing the authors tomorrow for episode #2 of Deep Papers. Paper: arxiv.org/abs/2212.14052 (New competitor to the Transformer architecture that scales ~linearly instead of quadratically with context length).

Is Microsoft bing doing an epic games store move?

Very very cool ideas ai.googleblog.com/2022/12/accele…

I tried out GPT4 for a real coding task. Based on impressions on Twitter and multiple papers, I was expecting to be blown away. And honestly... meh. 🧵1/

To make neural networks as modular as brains, We propose brain-inspired modular training, resulting in modular and interpretable networks! The ability to directly see modules with naked eyes can facilitate mechanistic interpretability. It’s nice to see how a “brain” grows in NN!

Reading up on (broadly speaking) relational learning, Dora and Glom. Why can't I find a single cross-reference between Geoffrey Hinton and Leonidas Doumas?

optogenetics, computational neuroscience, neuropixels, whole cell patch clamp

To support the labs SussexNeuro we have an open technical position for an #electronics specialist! The person will be working with me on developing #openhardware/#openscience tools for #neuroscience Initially for 2 years, but with the possibility of extension 1/2

It buffles me how people underestimate intelligence: 1966 - it was thought that vision was basic & could be solved over summer by one student. 2023 - Still no solution in sight except for a few relatively basic advancements here & there.

"News and books are vital for training next generation of artificial intelligence models, warn developers." But so is revenue sharing to keep those news & books coming. Where do they think the news & books come from?

Wow best description of using lllms.

Has anyone read Chomsky? Honestly, it seems like either scientists learnt to sensationalise knowingly, or nobody has actually read the hypothesis they are attacking...

Why is this in nature? These results have been around for at least a year. Not to mention: duh, theory predicts that.

👇 The AI world pretty much divides into two groups: those who understand why current machine learning techniques suck at outliers, and therefore struggle at things like driverless cars and high-level reasoning in unusual circumstances — and those who don’t. Those who don’t have

Something might be really wrong - because it works. By merely enumerating terms, you can find "new algorithms". For example, if you try to solve for `f` in `f 0010 = 1010 && f 1100 = 0010`, you will quickly find the optimal binary increment algorithm. It is the first match. Even

This is big.

I'm superbly happy that keeping up with ml and ai means learning from Toph, talking wolves and people who use the word "goon" in academic-grade posts.

Looks mighty cool, will try soon! arxiv.org/abs/2504.14216

You may be early or you may be late, but if you're doing AI, you will eventually be part of the wave.