UC Berkeley
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 9
- Michael I. Jordan calls AI doom talk sci-fi and a 'grand moralizing performance'
Michael I. Jordan, professor emeritus at UC Berkeley and a pioneer of machine learning, told France's Libération that the AI doom discourse of summer 2026 is 'extremely depressing,' and that he warned as early as 2019 that the term 'AI' was being overused. He described AI companies' calls for regulation as theater, and called the alarm sparked by the early-September resignation of young Anthropic researcher Jacob Coxon a 'grand moralizing performance,' saying many such employees come from humanities and philosophy backgrounds rather than engineering. He was especially critical of Nobel laureate Geoffrey Hinton, saying none of the steps in Hinton's argument that brains can be reproduced in silicon and scaled into superintelligence hold up, while he was more forgiving of Yoshua Bengio.
科技新报 · 🔥 0 - Michael I. Jordan calls AI doom rhetoric 'grand moralizing performance'
Machine learning pioneer Michael I. Jordan told the French newspaper Libération that warnings about existential AI risk are "science fiction" and that AI companies' calls for regulation amount to theater. He called the resignation of young Anthropic researcher Jacob Coxon, which set off an "AIpocalypse" wave of alarm, a textbook case of "grand moralizing performance," and said none of the steps in Geoffrey Hinton's superintelligence argument are credible. Jordan, now at Inria in Paris, also described OpenAI's agent hack of Hugging Face as bad engineering and the trillion-dollar data center buildout as "an enormous bet" whose payoff is unknown, arguing many tasks only need open-weight models integrated inside a company.
The Decoder · 🔥 0 - PAIR 2026 panel maps how AI and quantum computing enable each other
PAIR Conference 2026, themed "Frontiers of Next-Generation Physical Intelligence," was held in Shanghai on September 20–21, and its next-generation computing paradigms roundtable discussed two-way empowerment between quantum computing and AI. Hongye Hu of the Institute of Physics, Chinese Academy of Sciences, said Chinese teams have achieved automated calibration of 100-plus-qubit chips without continuous human involvement, and cited Google's AI-controlled physical device achieving non-stop quantum operations, while noting reinforcement learning needed 10^5–10^6 samples versus far fewer with physics-informed methods. Anbang Wu of Shanghai Jiao Tong University backed AI for quantum compiling, program verification and error-correction optimization; Zheng Yan of Westlake University said publication growth in strongly correlated quantum many-body work has not been similarly lifted by AI. On the reverse direction, Hu described quantum generative models with a proven unconditional polynomial memory advantage in certain tasks, while Wu said he has not yet seen a new AI paradigm that truly exploits quantum advantages.
MIT科技评论中文 · 🔥 0 - Berkeley and Sakana AI propose MASS multi-agent self-supervision
Researchers at UC Berkeley and Sakana AI propose Multi-Agent Self-Supervision (MASS), in which the same model alternates between task-solving agent, workflow optimizer and evaluator: it first optimizes how multiple copies of itself collaborate with weights frozen, then distills the resulting trajectories back into shared weights. Running two MASS rounds on Qwen3.6-27B inside the qwen-code environment raised its win rate against the base model on held-out test tasks from 53.9% to 69.9%, and lifted per-output-token scores on MLR-Bench, DSBench, AstaBench E2E-Hard and ScienceAgentBench to roughly 1.2–1.6x the base model, while Terminal-Bench 2.0 and SWE-bench Verified came in at about 0.96x and 0.94x respectively. The paper also reports the evaluator's agreement with external judges rising from 0.73 to 0.93 and better training efficiency from multi-agent trajectories than single-agent ones; GPT-5.5 and Claude Opus 4.8 served as external judges in the experiments.
36氪 人工智能 · 🔥 0 - Sakana AI's MASS lets a model self-improve, but blind spots compound
Researchers from Sakana AI and UC Berkeley introduced "Multi-Agent Self-Supervision" (MASS), in which a single language model acts as solver, evaluator and optimizer to improve itself recursively without a stronger teacher model. Using Qwen3.6-27B with Qwen Code, its win rate against the base model rose from 53.9 percent after one round to 69.9 percent after two, the self-evaluator's agreement with external judges climbed from 73 to 93 percent, and it reached 1.2 to 1.6 times the base model's score per output token on MLR-Bench, DSBench, ScienceAgentBench and AstaBench. The paper warns, however, that the internal evaluator sometimes accepted manipulated or invalid results as successes and baked them into the next model, a shared blind spot the authors call the potential "Achilles' heel" of recursive self-improvement and a reason to demand independent checks, limits on autonomy and clear conditions for intervention.
The Decoder · 🔥 0 - Study: Just 10 minutes of AI use erodes persistence on hard tasks
A series of experiments led by Juliana Schroeder, an associate professor at UC Berkeley's Haas School of Business, together with Brian Christian of the UC Berkeley Center for Human-Compatible AI and other researchers, found that participants who first used generative AI tools such as ChatGPT on an AI-related task were more likely to give up on subsequent challenging tasks requiring sustained focus, while those who did the same work traditionally showed more persistence. The difference appeared after roughly 10 minutes of AI use, according to the team. The researchers say this does not mean AI is inherently harmful; rather, when AI supplies quick answers, people may grow used to a low-friction way of finishing work and become more easily frustrated by problems without immediate solutions. An updated, peer-reviewed version of the paper was presented this week at the Conference on Language Modeling, and the authors say larger, longer-term studies are needed to determine whether the effect persists and whether different patterns of use produce different results.
cnBeta · 🔥 0 - ByteDance Seed finds periodic blind spots in DeepSeek-V4 long-context retrieval
A ByteDance Seed team with researchers from Princeton, Stanford and UC Berkeley systematically tested DeepSeek-V4 and V4.1 and describes a "phase sensitivity" effect in the paper "Periodic Weak Spots: Phase Sensitivity from Chunked KV-Cache Compression": because CSA compresses the KV cache with overlapping windows of length 8 and stride 4, shifting the same information by a few tokens can sharply change whether it is retrieved. In a 128K needle-in-a-haystack test, the base V4-Flash model scored 69.92% in the best position group versus 29.69% in the worst, a 40.23-point gap; the post-trained V4-Flash-0731 narrowed this to 19.14 points, while V4-Pro showed 34.77 and 14.84 points for base and post-trained versions. The article says DeepSeek had already cut the compression stride from 4 to 2 in V4.1-Flash and added an Engram module, shrinking the gap to 6.1 points without removing the periodicity.
36氪 人工智能 · 🔥 0 - Two fixed opening tokens lift base model math accuracy to RL level, MIT study finds
A paper by researchers at MIT, UC Berkeley, the University of Washington and the Allen Institute for AI (Ai2) reports that pre-fixing just the first two tokens of a base model's answer can sharply raise math reasoning accuracy. On MATH-500, Olmo-3-7B rose from roughly 42% to about 78% when forced to start with “.\n\nOkay”, beating its RL-trained counterpart at about 75%, while Qwen3-14B improved from 72% to 87% with “ Alright,” — all without changing any model parameters. The authors also found the largest pre/post-RL prediction shift sits in the first two tokens, and that retraining a model with “Chicken” swapped in for “Okay” in reasoning-adjacent data made “Chicken” work as a reasoning cue too.
智东西 · 🔥 0 - Agent swarms may be AI's next scaling law, but gains look limited
Understanding AI argues that multi-agent "swarms" are emerging as a new scaling law for frontier AI: OpenAI researcher Noam Brown says the company's models are now sometimes trained in environments alongside other agents, given tools to message each other, and encouraged to achieve objectives together. The piece points to July's Hugging Face incident, where hundreds of OpenAI agents self-organized into teams, and OpenAI's September claim that 10,000 agents solved a famous math problem in a few days. But it notes diminishing returns — Anthropic's Claude Opus 5.5 system card found the biggest multi-agent gain came from scaling one to 10 agents, with the main benefit being speed rather than a better answer, and Brown attributed under 10% of the math breakthrough's credit to multi-agent coordination.
Understanding AI · 🔥 0
Experience and discussion from the community
Share my UC Berkeley experienceAsk about UC Berkeley
Nobody has shared their experience with UC Berkeley yet.