I am not a lawyer, but this seems maybe important.

Ataraxos, an AI for the board wargame Stratego, establishes a design pattern for reinforcement learning and search that is effective under large amounts of hidden information, a longstanding desiderat...

Retatrutide is a triple-receptor agonist of glucose-dependent insulinotropic polypeptide, glucagon-like peptide-1, and glucagon receptors. In this phase 3, randomized, double-blind trial, we assign...

Complex and self-aware personages deserve a shot at dignity. Let me show you one.

One life, drawn at random from all who have ever lived.

Learn how MaxText reproduced Ai2’s OLMo 3 7B on Google Cloud TPUs, matching PyTorch GPU benchmarks across pre-training with up to 57.4% MFU.

Stress-testing frontier AI in leaky labs

Ramez Naam gives a skeptic’s take on Recursive Self-Improvement.

Post-training quantization (PTQ) is widely used to deploy large language models efficiently, but its effect on reasoning models is not well understood. Across math, coding, and science QA, we find tha...

YouTube video by S3 | Science, Startups, & Stories

We found a clear internal signal in models that accompanies reward hacking, and built probes that detect it — enabling efficient, real-time detection of reward hacking at scale.

YouTube video by Atomic Frontier

Discover ARGUS, Pilgrim's advanced technology platform for national security and defense applications.

Vision Language Models (VLMs) offer the exciting possibility of processing text as rendered images, bypassing the need for tokenizing the text into long token sequences. Since VLM image encoders map f...

Shop Photon Matrix laser mosquito air defense systems—AI, LiDAR and radar-powered mosquito protection for homes, patios, camping and outdoor spaces, with no chemical spray.

YouTube video by Ziroth

Give your coding agents a shared memory. They can read each other's decisions, reuse good ideas, and recover the reasoning when a session ends.

Scaling laws predict how loss decreases with increases in computation. We show, contrary to conventional wisdom, that architectural interventions can modify scaling exponents in pre-training, leading ...

This is the fourth in a series of informal research updates from the Google DeepMind Language Model Interpretability team, in interpretability and ad…