Tensors are instrumental in physics, machine learning and even biology. Einstein once begged a friend to help him understand them, fearing he was going mad. Joseph Howlett explains how they work: quantamagazine.org/the-geometric-…

Following our Concordia Contest held at #NeurIPS2024, we are excited to announce Google DeepMind's release of Concordia v2.0 for multi-agent evaluations. Watch the tutorial with Alexander (Sasha) Vezhnevets: youtube.com/watch?v=2FO5g6…

It’s rare for competitors to collaborate. Yet that’s exactly what OpenAI and Anthropic just did—by testing each other’s models with our respective internal safety and alignment evaluations. Today, we’re publishing the results. Frontier AI companies will inevitably compete on

Our next seminar with Zhijing Jin ✈️ Dec 2@EurIPS, Dec 5-8@NeurIPS is just around the corner. Join us this Thursday to explore findings on LLM cooperation based on two social simulations: GovSim and SanctSim. Link to register below.

A big part of our mission at Thinking Machines is to improve people’s scientific understanding of AI and work with the broader research community. Introducing Connectionism today to share some of our scientific insights.

Excited to share details on two of our longest running and most effective safeguard collaborations, one with Anthropic and one with OpenAI. We've identified—and they've patched—a large number of vulnerabilities and together strengthened their safeguards. 🧵 1/6

atlas.mitre.org/matrices/ATLAS

Can cooperative AI enhance democratic deliberation? Talks from our 2025 Cooperative AI Summer School exploring this are now available on YouTube. Links below.

.Richard Sutton, father of reinforcement learning, doesn’t think LLMs are bitter-lesson-pilled. My steel man of Richard’s position: we need some new architecture to enable continual (on-the-job) learning. And if we have continual learning, we don't need a special training

Introducing Tinker: a flexible API for fine-tuning language models. Write training loops in Python on your laptop; we'll run them on distributed GPUs. Private beta starts today. We can't wait to see what researchers and developers build with cutting-edge open models!

Desde aquí le mando un abrazo fuerte a Rosa. Espero que para cualquier cosina que me escriba . Era difícil. #Pasapalabra . Contiene la ñ!

Regalazo que me acaba de hacer Gema Parreño 😍😍😍😍 Recuerdo de haberlo petado en Essen, brutal. Ya estamos en Móstoles, mañana rumbo a Sevilla 💪💪

How can AI help us make progress on human cooperation challenges? Read our recent blogpost capturing the discussions on AI-facilitated human cooperation from our Cooperative AI Summer Retreat. Link below.

After reading many of the replies, we would like to issue a few clarifications: - we cannot extract training data from the model using our method - LLMs are not injective w.r.t. the output text, that function is definitely non-injective and collisions occur all the time -

Buen primer juego post essen!

Hoy toca juegazo! Gracias gracias gracias Llama Dice ! Nos vemos prontoo

We disrupted a highly sophisticated AI-led espionage campaign. The attack targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We assess with high confidence that the threat actor was a Chinese state-sponsored group.

Gemini 3! This is our most intelligent model that brings any idea to life. 😻 This is the best model in the world, by a crazy wide margin! Aside from a huge increase across the absolutely everything, look at its coding capabilities and quality of aesthetics and fidelity.

Attending #NeurIPS2025 NeurIPS Conference? We’ve highlighted some of the most exciting cooperative AI research papers to read en route:

Come find us at #NeurIPS2025’s poster session today! Our Research Analyst Chandler Smith will present ‘Evaluating Generalization Capabilities of LLM-Based Agents in Mixed-Motive Scenarios Using Concordia’. More details below: