Neural networks

Moritz Hardt, author of a new book on benchmarks, believes these “standardized tests” for artificial intelligence are fundamentally broken — but not unfixable.

A mathematical mechanism enables AI models to process language at scale, undergirding the current AI revolution. Our columnist asks: Why is attention the key?

Computer scientist Melanie Mitchell discusses why artificial intelligence doesn’t “think” or “reason” like humans, and how we can create better methods for measuring machine cognition.

The last decade has seen vast improvements in humanoid robots, but graduating to widespread use might require going back to the fundamentals.

A $1 million prize awaits anyone who can show where the math of fluid flow breaks down. With specially trained AI systems, researchers have found a slew of new candidates in simpler versions of the problem.

Is the inside of a vision model at all like a language model? Researchers argue that as the models grow more powerful, they may be converging toward a singular “Platonic” way to represent the world.

The V-JEPA system uses ordinary videos to understand the physics of the real world.

Naomi Saphra thinks that most research into language models focuses too much on the finished product. She’s mining the history of their training for insights into why these systems work the way they do.

Fundamental technique lets researchers use a big, expensive “teacher” model to train a “student” model for less.