happy

Robotics and AI

[Sakana AI @SakanaAILabs]

「AIは訓練するのではなく、自ら”育つ”ようにすべきだ」

Sakana AIのリサーチャー、@sebastianrisi がポッドキャスト @EyeOn_AI に出演。進化的手法でニューラルネットワークを構築するNeuroevolution手法の概要、継続学習や人工生命（ALife）研究の今について語りました。

https://t.co/fOlAyrANUg

[Luiz Pessoa @PessoaBrain]

𝗢𝗻 𝗯𝗶𝗼𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗮𝗻𝗱 𝗮𝗿𝘁𝗶𝗳𝗶𝗰𝗶𝗮𝗹 𝗰𝗼𝗻𝘀𝗰𝗶𝗼𝘂𝘀𝗻𝗲𝘀𝘀: 𝗔 𝗰𝗮𝘀𝗲 𝗳𝗼𝗿 𝗯𝗶𝗼𝗹𝗼𝗴𝗶𝗰𝗮𝗹 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹𝗶𝘀𝗺
Excellent paper discussing how considering biology is important to understand consciousness.
https://t.co/f5bizhkQCy

[R here @UvgtdERt4PWmcjF]

ระบบกระสุนปืนของต้น Oxalis ยิงเมล็ดพืชออกมาอย่างระเบิดขณะที่ฝักแห้งได้ที่ ทันทีที่ถูกสัมผัส

[eLife - the journal @eLife]

⛓️ October’s most-read Structural Biology paper challenges textbook views of synapses, revealing a complex ‘synaptoplasm’ instead of a dense post-synaptic layer: https://t.co/5m0rvmW1Yx

Have a paper people should see? See what our Editors look for: https://t.co/jYrVfaJKT4

[Earl K. Miller @MillerLabMIT]

The functional role of oscillatory dynamics in neocortical circuits: A computational perspective
https://t.co/Fr506aMlK3
#neuroscience

[Keith Sakata, MD @KeithSakata]

Neuroplasticity = the brain’s ability to change.

But plasticity itself isn't “good” or “bad.”

In a bad environment:
• High plasticity → rapid learning of the wrong lessons
• Low plasticity → can't unlearn them

Here's how to unlock your plasticity 🧵

[Wonder of Science @wonderofscience]

Isolated groups of neurons trying to connect to each other through micro-tunnels, captured in time-lapse with a microscope.

📽: Dr. Renaud Renault

[Chenhao Li @breadli428]

🐶Motion imitation should never limit itself to simply replaying the motions in the dataset on flat terrains.

In our #CoRL2025 work Motion Priors Reimagined, we extend motion tracking to complex locomotion and navigation tasks, to make skills "useful".

🔗https://t.co/X4QZgX5q2J

[Matt Dancho (Business Science) @mdancho84]

Principal Component Analysis (PCA) is the gold standard in dimensionality reduction.

But PCA is hard to understand for beginners.

Let me destroy your confusion:

[Jyo Pari @jyo_pari]

For agents to improve over time, they can’t afford to forget what they’ve already mastered.

We found that supervised fine-tuning forgets more than RL when training on a new task!

Want to find out why? 👇

[Cliff Pickover @pickover]

Mathematics. A Universe of Linear Algebra.

Welcome to Matrix World.

Source: https://t.co/SL6LvYPF2U

[François Rozet @FrancoisRozet]

Does a smaller latent space lead to worse generation in latent diffusion models? Not necessarily! We show that LDMs are extremely robust to a wide range of compression rates (10-1000x) in the context of physics emulation.

We got lost in latent space. Join us 👇

[Simone Scardapane @s_scardapane]

*Alice's book got a (minor) upgrade!*

Thanks to the dozens of people who gave feedback, now with 1000% less typos and errors, a novel set of Colab lab sessions, and a brand-new CC-BY-SA license. 🙃

https://t.co/jdjs9Gor0v

[Tom Yeh @ProfTomYeh]

I still remember back in grad school. My friend in NLP used to show off, bragging that he had LSTM all figured out. I envied him. Fortunately, my field was Computer Vision. I could survive just knowing my SVMs. In 2024, the inventor of LSTM himself is finally back with the extension: xLSTM. Here's my Excel implementation. Not for the faint of heart. Download: https://t.co/77wgxlQNoR I guess I can brag to my NLP friend now. 😉

[Probability and Statistics @probnstat]

Autoencoders are neural networks that learn efficient data representations by encoding inputs into a lower-dimensional space and then reconstructing them. They’re widely used for dimensionality reduction, denoising, and feature learning.

https://t.co/HM3IlK1RdT

[Sam Earle @Smearle_RH]

We introduce PuzzleJAX, a benchmark for reasoning and learning. 🧩💡🦎

PuzzleJAX compiles hundreds of existing grid-based PuzzleScript games to hardware-accelerated JAX environments, and allows researchers to define new tasks via PuzzleScript's concise rewrite rule-based DSL.

[neurosock🧠Brain Chips🦾 @neurosock]

The brain uses orthogonal sub-dimensions in neural space as communication channels.

This is a great new paper using Neuropixels from ~6500 neurons on 8 cortical and deep regions in mice.

Simplifying the space helps a lot to understand the idea.

Here is my toy model and notes:

[Samuel Schmidgall @SRSchmidgall]

Our paper on autonomous scientific research is accepted to Findings of #EMNLP2025! 🎉

We introduce Agent Laboratory, a framework that accelerates scientific discovery by teaming human researchers with LLM agents.

[Sander Dieleman @sedielem]

Measuring information content in bits is very useful. Information theory made digital communication, cryptography and machine learning possible.

But information is not just a quantity: it also has a shape. (1/6)

[Kate Deyneka @katedeyneka]

The fastest way to get smarter is to read smarter people.

I went through @tbpn’s Metis List and pulled together blogs from some of the most influential people in AI. I’ve also added a few gems outside the list 🧵

I find the most value in blogs where people share applied insights and digest SOTA research. Share your favorite AI blogs in the comments!