Satoshi Kuroki

both Bio and AI/ Electrophysiology/ Imaging/ Genetic engineering/ Behavior/ Place formation/ Continual learning/ Oscillation/ Hippocampus/ RNN/ 仕組みが知りたい

[Satoshi Kuroki @sskd3]

Our paper (my hobby research) is out. Using an extended Kuramoto model—the classic minimal framework for collective synchrony—we investigated the relationship between Excitation-Inhibition balance and synchronous activity.
https://t.co/P6uWMsQ2vL

[Satoshi Kuroki @sskd3]

jev + looped transformer みたいなのってできるのかな⁈

[Satoshi Kuroki @sskd3]

論文が出ました～(趣味の方)。集団同期の数理モデルであるKuramotoモデルを用いて、興奮性-抑制性バランスと同期活動の関係を調べました。
https://t.co/P6uWMsQ2vL

[hardmaru @hardmaru]

New Paper: Continuous Thought Machines 🧠

Neurons in brains use timing and synchronization in the way that they compute, but this is largely ignored in modern neural nets. We believe neural timing is key for the flexibility and adaptability of biological intelligence.

We propose a new neural architecture, “Continuous Thought Machines” (CTMs), which is built from the ground up to use neural dynamics as a core representation for intelligence. By using neural dynamics as a first-class representational citizen, CTMs naturally perform adaptive computation.

Many emergent, interesting behaviors arise as a result: CTMs solve mazes by observing a raw maze image and producing step-by-step instructions directly from its neural dynamics. When tasked with image recognition, the CTM naturally takes multiple steps to examine different parts of the image before making its decision. This step-by-step approach not only makes its behavior more interpretable but also improves accuracy: the longer it “thinks,” the more accurate its answers become.

We also found that this allows the CTM to decide to spend less time thinking on simpler images, thus saving energy. When identifying a gorilla, for example, the CTM’s attention moves from eyes to nose to mouth in a pattern remarkably similar to human visual attention.

I think this work underscores an important, yet often lost, synergy between neuroscience and AI. While modern AI is ostensibly brain-inspired, the two fields often operate in surprising isolation. By starting with such inspiration and iteratively following the emergent, interesting behaviors, we developed a model with unexpected capabilities, such as its surprisingly strong calibration in classification tasks, a feature that was not explicitly designed for.

When we initially asked, “why do this research?”, we hoped the journey of the CTM would provide compelling answers. By embracing light biological inspiration and pursuing the novel behaviors observed, we have arrived at a model with emergent capabilities that exceeded our initial designs. We are committed to continuing this exploration, borrowing further concepts to discover what new and exciting behaviors will emerge, pushing the boundaries of what AI can achieve.

[Satoshi Kuroki @sskd3]

ついに来たね(何かの真似)
AI co-scientist

試してみ��い

[K.Ishi@生成AIの産業応用 @K_Ishi_AI]

Googleは新しく、科学研究をアシストする「AI co-scientist」を発表した。

これが面白いのは、AIが自分自身で科学に関するディベートを行い、仮説をブラッシュアップしていくところだ。

計算時間をかければかけるほど質が良くなり、ある地点で人間の専門家の仮説を凌駕し、さらに上がる。

[Satoshi Kuroki @sskd3]

危険な混雑の検出に役立つみたいに書いてあるけど、これは普通に人の密度で判断すればいい気もする笑

[Satoshi Kuroki @sskd3]

ヒトがたくさん集まって混み具合が閾値を超えると渦を巻き始める。人が移動しようとする力と、密度が上がると人と人との間で生まれる抵抗の���うな相互作用（押し合いを逃れる力?）で渦を巻く。面白い。
https://t.co/lXyi48qE1U

[Satoshi Kuroki @sskd3]

大変勉強になりました
強化学習でLLMを学習させる方法(DeepSeek-R1の解説記事)
https://t.co/98iw3Wj7Hb

[Satoshi Kuroki @sskd3]

Red-shifted Ach GRAB sensor
https://t.co/mU6jtZltlc

[Satoshi Kuroki @sskd3]

ファスデングで若返る仕組み
��して腸内細菌が絡むらしい
https://t.co/qupeqLWDa2

[Andy Peters @flickerfusion]

Neuropixels Trajectory Explorer: plan trajectories (now including Neuropixels 2.0), see position in real-time experiment with New Scale MPM integration.

Available in MATLAB and as standalone:
https://t.co/ynkHaEfwmE

[Yoshikai @yoshikai_man]

Claude3 Opusに論文に載っているモデルのアーキテクチャ図を渡してPytorchでコード書いてもらった。GPT-4より出力が圧倒的に速いのがいい

元の論文はこれです
https://t.co/AaNYvz4Hke

[Satoshi Kuroki @sskd3]

ChatGPTが出た直後の開発止めよう論争は何だったのか…

[AGIラボ @ctgptlb]

ClaudeがGPT-4超えを達成した。GeminiもGPT-4を超えた。でもOpenAI が GPT-4 をリリ��スしたのは、なんと去年の3月。つまり、ちょうど丸一年間も経った。この間にOpenAIでGPTのモデルに大幅な性能向上がなかったはずがない。今週、GPT-5が出るかもしれない‥‥！

[Satoshi Kuroki @sskd3]

スクリーンサーバー(久々に言った)に使いたい

[Koen van Gilst @vnglst]

Saw this on Twitter somewhere and I had to recreate it. This is my coding happy place. 🪷

[kohei @koheiichi]

おお、transformerの時間発展と蔵本モデルや合意形成モデルなどの集団動力学モデルのアナロジーに着目し、transformer内にできるクラスター構造について議論している。

非線形物理のラボにいた身からするとめちゃくちゃ面白い

[Satoshi Kuroki @sskd3]

Q*, 名前的にA*アルゴリズム＋DQNっぽいなとぼんやり思ってたらホントにそれっぽい

[goto @goto_yuta_]

話題のQ*とやらの元ネタ疑惑論文。
タイトル見るとまじでそんな気がしてくるな笑

[Satoshi Kuroki @sskd3]

今年もタイムラインがSfNで溢れる季節がやってきた

[Hongyu_Chang @hongyu_chang]

For those exploring animal VR setup check this out: With @misaacson, we present the first miniaturized VR headset, MouseGoggles, for mouse neuroscience and behavior (< $200 in parts💰, open-source, highly immersive)🧵1/2
#VR @researchsquare @SN_Lab 🧠🐭🥽
https://t.co/0NteFHmBzM

[Satoshi Kuroki @sskd3]

我が家にもCode interpreterがやってきた