Tim

⚛️🚁🛰️🚀⛰️
Physics | Space | Novelty Search | Type-2 Fun
Ask better questions.

[Tim @tims457]

Opus 5 is so bad. I had to cancel my subscription. I do hope @AnthropicAI fixes @claudeai

[Tim @tims457]

@bcherny It might score high, but it's awful to work with. It's the worst communicator of any of your models.

[Tim @tims457]

Really nice feature in @claudeai. Automatic daily lessons in whatever you're trying to learn about.

[Anthropic @AnthropicAI]

New Anthropic research: A global workspace in language models.

Of everything happening in your brain right now, only a tiny fraction is consciously accessible—thoughts you can describe, hold in mind, and reason with.

We found a strikingly similar divide inside Claude.

[Tim @tims457]

Predictive Coding Networks and Inference Learning: Tutorial and Survey
https://t.co/sz4CCnJX1o

[Tim @tims457]

@AaronBMacLean This would be a good episode topic.

[Tim @tims457]

"Experiments in eutectic ice – a slushy mixture of water and salts that has been shown to promote polymerase activity and be a plausible environment on prebiotic Earth – revealed that QT45 could catalyse the synthesis of its complementary strand and also copy itself."

https://t.co/MrHlnhdTKy

[Machine Learning Street Talk @MLStreetTalk]

What if life itself is just a really sophisticated "computer program" that "wrote itself" into existence?

Blaise Agüera y Arcas @blaiseaguera presenting at @ALifeConf on how self-replicating programs spontaneously emerge from random noise.

This is the most technical public walkthrough of his "What is Life?" and "What is Intelligence?" books that exist on the internet.

He covers the BFF experiments, the phase transition math (Smoluchowski coagulation, Lotka-Volterra dynamics, eigenvalue analysis), and makes the case that symbiogenesis (and NOT mutation), is the primary engine of evolutionary novelty.

Whether or not you buy the full argument (the substrate is still carefully engineered, the mapping from toy computational systems to real biochemistry is non-obvious, and 'life was intelligent from the start' is load bearing.

That said -- the experimental results are genuinely really cool and the theoretical framework connecting abiogenesis to gelation phase transitions is fascinating research.

[Tim @tims457]

@M_McDonough Cool! Reminds me of this project I worked on a few years ago.
https://t.co/AJ6xgfj7Dc

[Tim @tims457]

Built my own simple news graph inspired by @EpsilonTheory's narrative machine.

https://t.co/DdNqsSodcw
(free instance so it can be slow to load)

[Tim @tims457]

@demishassabis How about a discussion about this with both of you on @MLStreetTalk ?

[Tim @tims457]

https://t.co/MRt1j4ukHD

[a16z @a16z]

The a16z infra team's book recommendations (heavy on sci-fi)

https://t.co/uCrNiluBsK

[Tim @tims457]

The book list thread

[Tim @tims457]

https://t.co/9HpY3d82wo

[Michael Nielsen @michael_nielsen]

A few suggestions I made for the @AsteraInstitute library

[Tim @tims457]

@aidan_mclau Is this a good example of a systems 1 course?
https://t.co/Z0bfhYuWss

[Google DeepMind @GoogleDeepMind]

We’re helping to unlock the mysteries of the universe with AI. 🌌

Our novel Deep Loop Shaping method
published in @ScienceMagazine could help astronomers observe more events like collisions and mergers of black holes in greater detail, and gather more data about rare space phenomena. 🧵

[Tim @tims457]

https://t.co/sBkRmMzdu9

[Machine Learning Street Talk @MLStreetTalk]

AI is so smart, why are its internals 'spaghetti'? We spoke with @kenneth0stanley and @akarshkumar0101 (MIT) about their new paper: Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis. Co-authors: @jeffclune @joelbot3000

[Sakana AI @SakanaAILabs]

Introducing The Darwin Gödel Machine: AI that improves itself by rewriting its own code

https://t.co/wEEB4LGPr0

The Darwin Gödel Machine (DGM) is a self-improving agent that can modify its own code. Inspired by evolution, we maintain an expanding lineage of agent variants, allowing for open-ended exploration of the vast design space of such “self-improving” agents.

Modern agentic systems, while powerful, remain static—once deployed, their intelligence remains fixed. We believe continuous self-improvement is key to the development of stronger AI capabilities. Our Darwin Gödel Machine is built from the ground up to enable AI systems that can learn and evolve their own capabilities over time, just as humans do.

On SWE-bench, DGM automatically improved its performance from 20.0% to 50.0%. Similarly, on Polyglot, the DGM increased its success rate from an initial 14.2% to 30.7%, significantly outperforming representative hand-designed agents.

Learn more about our approach in our technical report: https://t.co/kDNWFgCI6C

This work was done in collaboration with Jeff Clune (@jeffclune)’s lab at UBC, and led by his PhD students Jenny Zhang (@jennyzhangzt) and Shengran Hu (@shengranhu), together with Cong Lu (@cong_ml) and Robert Lange (@RobertTLange).

Code: https://t.co/RcYLd22TB5

[Tim @tims457]

@ericweinstein How do you explain the trades? Trades would need to be real and make money most of the time. Madoff's claimed trades never happened.

[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.