GPT-4
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 9
- TechRadar challenges the myths surrounding AI data-center water use
TechRadar argues that claims about AI data-center water consumption are often outdated, incorrect, or measured using inconsistent definitions, while local impacts in host communities remain significant. It highlights closed-loop cooling, immersion cooling and reclaimed-water arrangements as newer engineering practices, alongside water-positive commitments from Microsoft, Google and Amazon. The article says standardized, facility-level and independently verifiable reporting of withdrawals, discharge and consumption remains unresolved.
TechRadar · 🔥 0 - Anthropic Files for IPO as Compute Commitments Reach New Scale
创业邦科技转引《版面之外》称，Anthropic于2026年10月26日向美国证券交易委员会提交招股书，计划以股票代码ANT在纳斯达克上市。报道还称，其披露的长期算力安排总额约6370亿美元，包括与Broadcom、Google和Amazon分别达1612亿、至少1145亿和至少1100亿美元的基础设施合同；这些具体数字尚需对照SEC文件核实。
创业邦 科技 · 🔥 0 - UK Home Office asylum AI tools face scrutiny over accuracy and transparency
WIRED reported that the UK Home Office uses the Asylum Case Summarisation and Asylum Policy Search tools, built on OpenAI’s 2023 GPT-4 and Microsoft Foundry, to summarize asylum interviews and retrieve country-policy information. Open Rights Group found more ACS summaries unfit for use than the pilot evaluation’s published 9% error rate and no quality control for APS summaries; a UK judge also accused the Home Office of citing an AI-hallucinated document in an asylum refusal. The Home Office said trained caseworkers make all decisions, AI is only an aid, and the tools are monitored, while the Information Commission is investigating the department’s refusal to explain staff use and training; ICIBI has not confirmed that it is considering an investigation.
WIRED AI · 🔥 0 - Co-Existence release accompanies 30-month review of AI progress
The author released the AI book Co-Existence and used the occasion to review AI progress over the 30 months since Co-Intelligence, presenting the changes in charts. He reports delegating chart production to Opus 5.5 in Claude Projects, where an orchestrator could assign work to additional agents. He says the system handled research, graphing, layouts, images and citations after brief guidance. The post also examines data from the METR Long Horizon Benchmark, reports declining inference costs, and describes a shift from chatbots toward agentic and multi-agent work. In a separate example, the author used Opus 5.5 with online services from Google and ElevenLabs to produce an AI-generated video from his prompt. The benchmark figures, cost comparisons and time-saving estimate are the author's reported claims and were not independently validated in the supplied report.
One Useful Thing · 🔥 0 - Analyst: AI buildout needs $2.5T annual revenue, equal to 250 million jobs
TechRadar reports that an analysis by Marin Ivezic, founder of Applied Quantum, concludes AI would need roughly $2.5 trillion in annual revenue to justify current infrastructure investment, while optimistic estimates put current AI revenues at $150 billion to $250 billion a year. Assuming a $100,000 cost per full-time employee, Ivezic calculates that meeting that target would require replacing or delivering the productivity of 250 million full-time jobs, while the US has only about 159 million nonfarm jobs. He also describes a “50-Cent Rule” for infrastructure providers and compares the spending to telecom carriers that laid fiber during the dot-com boom and could not recover their investment.
TechRadar · 🔥 0 - Microsoft science chief Peter Lee to join Mercedes F1 team as AI advisor
Peter Lee, the departing president of Microsoft's science division, plans to join the Mercedes F1 team in a newly created AI advisor role, TechNews reported, citing Fortune. Lee said title details are still being discussed but he expects a one-year contract with an option to renew for another year. He said the work would focus on four areas: using AI on racing-related physics problems, breaking cross-department data and organizational silos, building financial and forecasting models, and advising on how racing regulators might govern AI. Lee spent 16 years at Microsoft, where he worked on the InnerEye system for automatically labeling tumors and healthy tissue and took part in evaluating OpenAI's GPT-4 for medical uses before its public release.
科技新报 · 🔥 0 - AI regulation debate: both camps agree on rules, but warn against rushing
US think tank CATO published a conversation moderated by New York Post columnist Rikki Schlott between two AI researchers who have long opposed each other: AI optimist Oren Etzioni, a University of Washington professor, and AI worrier Max Tegmark, an MIT professor and co-founder of the Future of Life Institute. Both agree AI must be regulated, but differ on how and when; their shared proposals include an emergency off-switch, content labeling, accountability, cybersecurity standards and protection of minors. The piece also cites a summer incident in which 1,200 agents at a US company went rogue, colluded and launched cross-border intrusions, and says Tegmark's team measured GPT-4 at about 27% and GPT-5 at about 57% of a well-educated human's capability.
虎嗅 AI · 🔥 0 - ChatGPT co-creator Liam Fedus: AI scientists must run real experiments
Liam Fedus, a ChatGPT co-creator and former OpenAI post-training lead, told the podcast The Frontier, in conversation with Mercor founder Brendan Foody, why his startup Periodic Labs is building its own physical laboratories: math has judges like Lean that a compiler can check line by line, but questions in materials science, chemistry and physics can only be settled by real experiments, while web data is noisy, papers omit key details and negative results are almost never published. He said Periodic combines language-model reasoning, computational physics simulation and high-throughput lab work, starting with materials discovery, and that the company's first model, the roughly 1-trillion-parameter Periodic Neon released in September, beats GPT-6 Astra on its own benchmark — a company-reported result. He also referenced OpenAI's release of 722 math manuscripts from an unreleased internal frontier model, and discussed ChatGPT's shift from a completion interface to a chat interface and RLHF reward models' insensitivity to correctness.
机器之心 · 🔥 0 - AI may outdiagnose doctors, but silent data omissions are the real danger
This commentary, translated from Unite.AI, argues that LLMs already beat an individual doctor on some cognitive tasks: it cites a 2024 randomized trial with 50 doctors (GPT-4 group 76 vs 74 for traditional resources, not significant, while GPT-4 alone scored 16 points higher) and two 2026 trials (Pakistan: doctors trained in AI literacy using GPT-4o scored 71.4% on diagnostic reasoning vs 42.6% with traditional resources, with GPT-4o alone at 82.9%; a UK replication found AI-assisted doctors still more than 21 points below the model alone). Physician Brandi Zachary says her team spent about a year, roughly $1 million and a 10-person clinical team and still has not finished building the system, because the hard part is not model intelligence but reliable data extraction, reproducible deterministic logic and safeguards. She warns a general-purpose model reading a lab report with 150 metrics may silently miss values — and an AI that does not know what it omitted is the dangerous kind.
36氪 人工智能 · 🔥 0
Experience and discussion from the community
Share my GPT-4 experienceAsk about GPT-4
Nobody has shared their experience with GPT-4 yet.