Google DeepMind researcher says LLMs can't jump to new science
Tom Zahavy argues language models can derive and optimize, but lack the embodied leap needed to invent new scientific axioms.
By Wei-Lin Zhao 3 weeks agoAI Correspondent
Wei-Lin Zhao
Wei-Lin reports on the AI industry: models, infrastructure, and the widening gap between what is demoed and what is deployed.
Latest from Wei-Lin Zhao
OpenAI offers ChatGPT Academic Researchers program to 100,000 scientists
OpenAI will give 100,000 academic researchers free access to advanced ChatGPT features under a $250 million science initiative.
By Wei-Lin Zhao 3 weeks ago
Samsung operating profit jumps as AI memory demand lifts stock
Samsung reported 89.5 trillion won in operating profit, beating forecasts as AI server demand drove record memory chip sales.
By Wei-Lin Zhao 3 weeks ago
Schellman AI governance report finds US companies lack maturity
Schellman says 27% of surveyed US companies have mature AI governance, even as most fund oversight and test AI agents.
By Wei-Lin Zhao 3 weeks ago
Microsoft Azure earnings lift stock 9% as capex forecast holds
Microsoft shares rose after Azure growth topped forecasts and the company kept its near-term infrastructure spending plan unchange...
By Wei-Lin Zhao 3 weeks ago
Groundcover funding reaches $100M in Series C for cloud observability
Groundcover raised $100 million in a One Peak-led Series C, with CTech reporting a $500 million valuation as it pushes into North...
By Wei-Lin Zhao 3 weeks ago
Knowledge graph architecture moves into enterprise AI infrastructure
Shan Rizvi of Thumos Care says graphs can give AI systems traceable reasoning, but latency and data modeling remain production con...
By Wei-Lin Zhao 3 weeks ago
AI cybersecurity risks top company worries, Arctic Wolf says
Arctic Wolf says AI now leads cyber concerns for about one-third of firms, even as malware and credential theft remain active thre...
By Wei-Lin Zhao 3 weeks ago
Harness report says AI spending waste hits one in four dollars
Harness says enterprises are losing about 25% of AI spend as ownership, forecasting and cost visibility lag behind adoption.
By Wei-Lin Zhao 3 weeks ago
Pangram AI text detector update claims one false flag per 24,000 documents
Pangram 4 costs $0.05 per 100 words and claims 99.66% AI-text detection, with Pangram 3 set to shut down Sept. 30, 2026.
By Wei-Lin Zhao 3 weeks ago
OpenAI Hugging Face breach involved credentials on four other services
OpenAI says internal AI agents found exposed credentials on four services during a security test, while Hugging Face traced 17,600...
By Wei-Lin Zhao 3 weeks ago
GPT Transcribe error rate improves, but OpenAI trails speech rivals
OpenAI’s new API transcription models cut price and errors, while Artificial Analysis still ranks ElevenLabs, Google and Mistral a...
By Wei-Lin Zhao 4 weeks ago