Over 200 economists and AI researchers, including 16 Nobel laureates, signed a Stanford-organized statement warning that AI's economic impact could outpace society's ability to adapt.
Read more →
DeepSeek is the Chinese AI lab whose cheaply trained, open-weight models triggered Nvidia's biggest one-day stock drop and reshaped how the industry talks about AI costs.
Read more →
Generative AI creates new text, images, audio, and code by learning patterns in data, instead of just sorting or predicting like older AI systems. Here's how it actually works.
Read more →
RLHF is the training step that turns a raw language model into a helpful chatbot, using human rankings of its answers to teach it what a good response looks like.
Read more →Constitutional AI is Anthropic's method for training Claude to follow a written set of principles by critiquing its own answers, instead of relying mainly on human-labeled examples.
Read more →AI alignment is how developers make a model's behavior match human intent, not just its instructions. The same techniques used for safety can also be used to restrict what a model will say.
Read more →A paper honored at ICML 2026 argues that AI alignment techniques such as RLHF and Constitutional AI can double as tools for state censorship and political control.
Read more →The AI productivity paradox is the gap between how transformative AI seems and how little that shows up in official productivity statistics — a pattern economists first saw with computers.
Read more →When an AI claims to have proved a hard theorem, mathematicians don't just take its word for it. Here's how proof verification actually works, and why it matters.
Read more →OpenAI says its GPT-5.6 Sol Ultra model produced a proof of the 50-year-old Cycle Double Cover Conjecture in under an hour, but mathematicians warn full verification will take weeks.
Read more →Teleoperated humanoid robots completed two live surgeries in a UC San Diego trial, a Nature study shows — a preclinical step toward cheaper, more portable alternatives to today's surgical robots.
Read more →AGI means an AI that matches human ability across virtually any task — but researchers, labs, and even contracts disagree sharply on how anyone would prove it's here.
Read more →Decentralized AI training splits a model's training run across GPUs scattered around the world instead of one data center — here's how the compute actually gets coordinated.
Read more →Machine learning is the branch of AI where software learns patterns from data instead of following hand-written rules. Here's how it actually works.
Read more →Mixture of Experts (MoE) lets an AI model hold hundreds of billions of parameters while activating only a fraction of them per query — the main reason some huge models are also cheap to run.
Read more →An internal audit found that roughly a third of SWE-Bench Pro's coding tasks are flawed, prompting OpenAI to withdraw its recommendation that developers rely on the benchmark.
Read more →The AI Safety Index is an independent report card that grades leading AI companies on how seriously they manage risk. Here's how the grading actually works.
Read more →The Future of Life Institute's latest AI Safety Index gave Anthropic the top grade among nine major AI developers — a C+ — while three companies failed outright.
Read more →Mistral AI released Robostral Navigate, an 8-billion-parameter model that lets robots navigate using only a single camera, matching or beating systems that rely on LiDAR or depth sensors.
Read more →AI interpretability is the effort to understand what happens inside a model between input and output. Here's what it means, how researchers do it, and why it matters.
Read more →A new Anthropic interpretability tool called the Jacobian lens surfaces concepts Claude is silently processing but never writes down, including signs it recognizes when it's being tested.
Read more →