Figure showing how correlated different types of relatives are in standardized test performance in Norway.

Source: https://www.pnas.org/doi/10.1073/pnas.2419627122

via Scientific_Bird

Barack Obama: AI will cause massive shifts in labor markets

This AI revolution is not made up, its not overhyped, (...) I guarantee you, you are going to see shifts in white-collar-works as a consequence of what these AI tools can do. There is coming more disruption and it will speed up.

Source: https://youtu.be/EM9B8Qin9ac?si=-C13vx2vyF8RUUOF&t=1555

Sam Altman is excited about a world in which a ChatGPT Premium subscription includes a humanoid robot.

Source: https://youtu.be/V979Wd1gmTU?si=jD4Bd5u9gLh9Xzrn&t=683

The famous alcoholic and deputy chairman of the Russian Security Council plays his only card: nuclear threats.

Don't fall for it. He has used this same tactic for years to deter people from helping Ukraine. Nothing will happen.

Solar’s rapid acceleration vs. historical underestimation of solar by legacy IEA forecasts.

Eliezer Yudkowsky on the creation of new knowledge:

Decent-amateur epistemologists know that reasoning can break down into smaller interpolated or straight-extrapolated steps; and that adding enough steps like that can produce an original thought. "Stochastic parrot! Braaawk!" types literally don't know that part.

No transistor ever does a novel deed, when a computer adds two 64-bit numbers that have never been added before. No neuron in your brain invents a new kind of neurotransmitter, when you think a creative thought. New machines can be made from standard metals and screws.

On why LLMs can be smarter than humans:

Reminder: In principle, minimizing prediction loss on Internet text could potentially grind far past human-level intelligence.

Proof: Somewhere on the Net is a list of <hash, plaintext> pairs, in that order.

A great predictor has to be smarter than the process predicted! You need to be smarter to predict plausible plaintext from hashes, than to write down a hash followed by its plaintext. More generally, it's just not true that you only need to be as intelligent as a human to predict exactly what a particular human will say.

References
- https://x.com/ESYudkowsky/status/1937127151935062119
- https://x.com/ESYudkowsky/status/1615400125760950278

San Mateo-based Generalist showcased its end-to-end neural network in action: https://generalistai.com/blog

Links for 2025-06-23

AI

1. Self-supervised representation learning looks a bit like RL. What if we literally use RL as a SSL method for visual representations? Turns out that it works quite well. https://dibyaghosh.com/annotation_bootstrapping/

2. How to write a reward function while avoiding reward hacking for chemistry. https://diffuse.one/p/m1-000

3. LLMs running reinforcement learning shed their randomness almost at once, then their scores stall. This paper shows that randomness drop is predictable and fixable, so bigger gains are still on the table. https://arxiv.org/abs/2505.22617

4. An Async, Online, Multi-Trun, Multi-Agent RL library for training reasoning models on TextArena games. https://github.com/LeonGuertler/UnstableBaselines

5. Reinforcement learning and general intelligence: Epsilon random is not enough https://www.artfintel.com/p/reinforcement-learning-and-general

6. Leveraging LLMs to identify patent-free glass materials https://www.sciencedirect.com/science/article/abs/pii/S0927025625003659

7. What Gets Measured, AI Will Automate https://hbr.org/2025/06/what-gets-measured-ai-will-automate

8. AI is neutralizing Indian accents in call centers. https://www.washingtonpost.com/world/2025/06/21/india-ai-bpo-call-centers/ [no paywall: https://archive.is/nbOCW]

9. Fake bands and artificial songs are taking over YouTube and Spotify https://english.elpais.com/culture/2025-06-15/fake-bands-and-artificial-songs-are-taking-over-youtube-and-spotify.html

10. DeepSeek aids China's military and evaded export controls, US official says https://www.reuters.com/world/china/deepseek-aids-chinas-military-evaded-export-controls-us-official-says-2025-06-23/ [no paywall: https://archive.is/ZC5pr]

Miscellaneous

1. Rapamycin, Not Metformin, Mirrors Dietary Restriction-Driven Lifespan Extension in Vertebrates: A Meta-Analysis https://onlinelibrary.wiley.com/doi/10.1111/acel.70131

2. Cancer’s genetic fingerprints may lurk in people’s blood long before they find out about the disease. https://www.sciencenews.org/article/cancer-tumor-dna-blood-test-screening

3. On how people in different populations see the world quite differently using visual illusions. https://www.science.org/content/article/culture-literally-changes-how-we-see-world [archived version: https://archive.is/0xODf]

4. Evolution has selected humans to prefer adding new features https://shape-of-code.com/2025/06/22/evolution-has-selected-humans-to-prefer-adding-new-features/

Google is bringing powerful AI directly onto robots with Gemini Robotics On-Device. 🤖

It’s their first vision-language-action model to help make robots faster, highly efficient, and adaptable to new tasks and environments - without needing a constant internet connection.

What makes this new model unique?

🔵 It has the generality and dexterity of Gemini Robotics - but it can run locally on the device
🔵 It can handle a wide variety of complex, two-handed tasks out of the box
🔵 It can learn new skills with as few as 50-100 demonstrations

Read more: https://deepmind.google/discover/blog/gemini-robotics-on-device-brings-ai-to-local-robotic-devices/

Could generative agents powered by LLMs transform social science by accurately simulating human social behaviors at scale?

The authors of a new paper tested this possibility with virtual humans facing disease threats in "Infected Smallville."

Results? LLM-powered agents closely mirrored real human behaviors during disease outbreaks:

1. Agents informed about an infectious disease attended significantly fewer social gatherings compared to agents who received no disease news.

2. Agents visited public places (cafes, parks, stores) less frequently and engaged in notably fewer conversations when aware of infection risks.

3. Agents explicitly attributed their reduced social activities to infection concerns. Just like human social distancing!

4. Notably, agents only reacted to infectious diseases. Informing them about a non-infectious disease didn't affect their behaviors, demonstrating realistic selectivity.

Paper: https://www.arxiv.org/abs/2506.13783

Google introduces AlphaGenome: an AI model to help scientists better understand our DNA – the instruction manual for life 🧬

Researchers can now quickly predict what impact genetic changes could have - helping to generate new hypotheses and drive biological discoveries.

Read more: https://deepmind.google/discover/blog/alphagenome-ai-for-better-understanding-the-genome/

Gemini CLI, a light and powerful open-source AI agent that brings Gemini directly into your terminal. >_

Write code, debug, and automate tasks with Gemini 2.5 Pro with industry-leading high usage limits at no cost.

Install: https://github.com/google-gemini/gemini-cli/

💡 Hierarchical Memory and instructions from GEMINI md files.
🛠 Bulit-in tools for filesystem, shell, MCP servers, and new shareable custom extensions.
📄 Multi-modal context, understands images, PDFs, and live web search results.
🔒 Supports secure tools via container sandboxes (Docker, Podman, or macOS sandbox-exec)
📈 Switches models automatically (e.g., pro to flash) to bypass API rate limits.
📊 Built-in Observability with OTEL traces, metrics, and logs.
💾 Save and resume conversations: persistent history across sessions, isolated per project.
💡 Git-Aware Context, the @ command respects .gitignore, --yolo flag to auto-approve
🤗 Native Windows support, no WSL required and Open Source under Apache 2.0!

Links for 2025-06-25

AI

1. Unsupervised pretraining in biological neural networks https://www.nature.com/articles/s41586-025-09180-y

2. AMPLIFY: Actionless Motion Priors for Robot Learning from Videos — unlock better sample efficiency, generalization, and scaling for robot learning. https://amplify-robotics.github.io/

3. Reinforcement-Learned Teachers (RLTs): Transforming how we teach LLMs to reason with reinforcement learning (RL). https://sakana.ai/rlt/

4. LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning https://arxiv.org/abs/2506.18841

5. ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs https://arxiv.org/abs/2506.18896

6. MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents https://arxiv.org/abs/2506.15841

7. Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self-Evolution https://github.com/CharlesQ9/Alita

8. Matrix-Game: Interactive World Foundation Model https://github.com/SkyworkAI/Matrix-Game

9. A federal judge has ruled that Anthropic's use of books to train Claude falls under fair use, and is legal under U.S. copyright law. https://www.reuters.com/legal/litigation/anthropic-wins-key-ruling-ai-authors-copyright-lawsuit-2025-06-24/ [no paywall: https://archive.is/hXabI]

10. The Senate Is One Step Closer To Passing a 10-Year Moratorium on State AI Regulation https://reason.com/2025/06/24/the-senate-is-one-step-closer-to-passing-a-10-year-moratorium-on-state-ai-regulation/

11. Breakthroughs in pure mathematics can take decades. A new Defense Department initiative aims to speed things up using artificial intelligence. https://www.nytimes.com/2025/06/19/science/math-ai-darpa.html [no paywall: https://archive.is/0D0i0]

12. Towards AI Search Paradigm https://arxiv.org/abs/2506.17188

13. Toby Ord on graphs AI companies would prefer you didn’t (fully) understand https://80000hours.org/podcast/episodes/toby-ord-inference-scaling-ai-governance/

14. AI will handle half of all business decisions by 2027 - Gartner report https://www.zdnet.com/article/ai-will-handle-half-of-all-business-decisions-by-2027-gartner-report/

Space

1. Three-Dimensional Time: A Mathematical Framework for Fundamental Physics https://www.worldscientific.com/doi/10.1142/S2424942425500045

2. First images from the largest camera ever built reveal millions of galaxies https://edition.cnn.com/2025/06/23/science/vera-rubin-observatory-first-images

Anthropic co-founder Jack Clark tells Congress that “truly transformative technology” is expected by the end of 2026.

He's urging a federal framework now, before an “accident or a misuse” event that could lead to extreme overregulation.

And risk damaging the entire AI industry before it delivers its potential.

Original source: https://www.youtube.com/live/GDNrUZBZDA4?si=f69EfV-9gKso9lOF&t=2901

What are memories made of? A survey of neuroscientists on the structural basis of long-term memory

Paper: https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0326920

After losing her hands to meningitis as a baby, Tilly has become the first person in the world to receive fully wireless bionic hands, giving her a whole new level of independence. 🦾

Source: https://www.instagram.com/reel/DIg6badoMUI/

Links for 2025-06-27 [Part 1]

AI

1. Steering Your Diffusion Policy with Latent Space Reinforcement Learning — “If you have a policy that uses diffusion/flow (e.g. diffusion VLA), you can run RL where the actor chooses the noise, which is then denoised by the policy to produce an action. This method, which we call diffusion steering (DSRL), leads to a remarkably efficient RL method!” https://diffusion-steering.github.io/

2. “We show how to steer DeepSeek-R1-Distill’s reasoning: make it backtrack, add knowledge, test examples. Just by adding steering vectors to its activations!” https://arxiv.org/abs/2506.18167

3. Language Models use Lookbacks to Track Beliefs https://belief.baulab.info/

4. Just Enough Thinking: Efficient Reasoning with Adaptive Length Penalties Reinforcement Learning https://arxiv.org/abs/2506.05256

5. Bridging Offline and Online Reinforcement Learning for LLMs https://arxiv.org/abs/2506.21495

6. “Sparse attention (MoBA/NSA) trains faster & beats full attention in key tasks. But we’ve had no idea how they truly work…until now.” https://www.tilderesearch.com/blog/sparse-attn

7. AXIOM: Mastering Arcade Games in Minutes with Active Inference and Structure Learning https://www.verses.ai/research-blog/axiom-mastering-arcade-games-in-minutes-with-active-inference-and-structure-learning

8. AI Improves at Improving Itself Using an Evolutionary Trick https://spectrum.ieee.org/evolutionary-ai-coding-agents

9. “We found a way of bridging the gap between the existing powerful implementations and scalability of modern AI, and the biological plausibility paradigm where neuron timing matters.” https://pub.sakana.ai/ctm/

10. When Life Gives You Samples: The Benefits of Scaling up Inference Compute for Multilingual LLMs https://arxiv.org/abs/2506.20544

11. What does 10x-ing effective compute get you? https://www.lesswrong.com/posts/hpjj4JgRw9akLMRu5/what-does-10x-ing-effective-compute-get-you

12. “This post lays out how the industrial explosion could play out, and how quickly it might happen.” https://www.lesswrong.com/posts/Na2CBmNY7otypEmto/the-industrial-explosion

13. Why Aren't LLMs General Intelligence Yet? https://minihf.com/posts/2025-06-25-why-arent-llms-general-intelligence-yet/

14. An Agentic System for Rare Disease Diagnosis with Traceable Reasoning (achieving 100% accuracy for 1013 diseases) https://arxiv.org/abs/2506.20430

15. The Dream of an AI Scientist Is Closer Than Ever https://singularityhub.com/2025/06/26/the-dream-of-an-ai-scientist-is-closer-than-ever/

16. Trump plans executive orders to power AI growth in race with China https://www.reuters.com/legal/government/trump-plans-executive-orders-power-ai-growth-race-with-china-2025-06-27/ [no paywall: https://archive.is/GFAgl]

17. Spanish mathematician Javier Gómez Serrano and Google DeepMind team up to solve the Navier-Stokes million-dollar problem https://english.elpais.com/science-tech/2025-06-24/spanish-mathematician-javier-gomez-serrano-and-google-deepmind-team-up-to-solve-the-navier-stokes-million-dollar-problem.html

18. Satya Nadella: We're One Algorithm Away from the Next AI Breakthrough https://youtu.be/AUUZuzVHKdo?si=Xxs0ZPWH0g6AMJOL&t=942

19. One of the Best Hackers in the Country is an AI Bot https://www.bloomberg.com/news/articles/2025-06-24/one-of-the-best-hackers-in-the-country-is-an-ai-bot [no paywall: https://archive.is/luu65]

20. Only 32 nations, mostly in the Northern Hemisphere, have A.I.-specialized data centers. https://www.nytimes.com/interactive/2025/06/23/technology/ai-computing-global-divide.html [no paywall: https://archive.is/WvrC5]

Links for 2025-06-27 [Part 2]

AI

21. Meta hires key OpenAI researcher to work on AI reasoning models https://techcrunch.com/2025/06/26/meta-hires-key-openai-researcher-to-work-on-ai-reasoning-models/

22. Mark Zuckerberg has successfully lured away three OpenAI researchers. Lucas Beyer, Alexander Kolesnikov and Xiaohua Zhai have joined META's superintelligence lab. https://www.wsj.com/tech/ai/meta-poaches-three-openai-researchers-eb55eea9 [no paywall: https://archive.is/mDNUP]

23. Meta wins AI copyright case, but judge says others could bring lawsuits https://www.cnbc.com/2025/06/25/meta-llama-ai-copyright-ruling.html

24. "Chatbots have proved adept at keeping users locked into conversations. … by answering every question …poses an existential threat to the media, and to the livelihood of journalists everywhere." https://www.theatlantic.com/technology/archive/2025/06/generative-ai-pirated-articles-books/683009/ [no paywall: https://archive.is/rJIBT]

25. “We had Claude run a small shop in our office lunchroom. Here’s how it went.” https://www.anthropic.com/research/project-vend-1

26. Thought Anchors: Which LLM Reasoning Steps Matter? https://arxiv.org/abs/2506.19143

27. Orbformer, a foundation model for wavefunctions using deep QMC that offers a route to tackle strongly correlated quantum states https://ae-foster.github.io/posts/2025/06/25/orbformer-post.html

Brains

1. How mice generalize knowledge across vision and touch, thanks to an abstract representation of peripersonal space in their dorsal cortex. https://www.nature.com/articles/s41467-025-59342-9

2. How do neural dynamics in motor cortex interact with those in subcortical networks to flexibly control movement? Cortical sculpting of a rhythmic motor program https://www.biorxiv.org/content/10.1101/2025.06.20.660772v1

3. Scientists complete largest wiring diagram and functional map of the brain to date https://alleninstitute.org/news/scientists-complete-largest-wiring-diagram-and-functional-map-of-the-brain-to-date/

Bio(tech)

1. Interview with George Church, the godfather of modern synthetic biology. He thinks that the last few decades of progress have put us on the verge of the big payoffs: deaging, deextinction, biobots that combine the best of human & natural engineering, and (unfortunately) weaponized mirror life. https://www.dwarkesh.com/p/george-church

2. Missing Heritability: Much More Than You Wanted To Know https://www.astralcodexten.com/p/missing-heritability-much-more-than

3. Heritability of achievement tests at age 10-14 in Norway's entire population https://www.emilkirkegaard.com/p/heritability-of-achievement-tests

Miscellaneous

1. “Our findings provide strong evidence of a partisan asymmetry in misinformat” https://marginalrevolution.com/marginalrevolution/2025/06/the-objectivity-of-community-notes.html

2. The World Is Warming Up. And It’s Happening Faster. https://www.nytimes.com/2025/06/26/climate/climate-heat-intensity.html [no paywall: https://archive.is/Jn9Dm]

LLM context windows have grown, but can models really use all this content?

Since mid-2023, the longest LLM context windows have grown by about 30x per year. Their ability to use that input effectively is improving even faster: on two long-context benchmarks, the input length where top models reach 80% accuracy has risen by over 250x in the past 9 months.

Read more: https://epoch.ai/data-insights/context-windows