#数据/训练

Former OpenAI researcher says Nvidia Blackwell fingerprints appear in six Chinese AI models

1-flash. The method exploits the random seed pattern PyTorch uses to initialize weights, which reflects a GPU's core count: 148 cores for B200/B300 and 152 for GB200/GB300, versus 132 for Hopper H100/H800. DePue says he reproduced the results for DeepSeek-V4-Flash and Kimi K3 on real Blackwell chips, published verification code on GitHub, and that the method correctly identified all 24 control models with known hardware.

Same event; the pick carries“Forensic analysis links six Chinese frontier models to export-controlled NVIDIA Blackwell GPUs”

How AI Is Helping Care Teams Get Patients the Right Care, Sooner

Dyania Health's Synapsis AI clinical reasoning engine reads every note, report and lab result in a patient's record, connecting findings years apart to flag which patients need attention and why, with each signal traceable to its source. It runs on NVIDIA-accelerated computing inside a health system's own firewall, so patient data never leaves the hospital. Cleveland Clinic used it to review records of more than 7,800 aortic valve replacement patients from 2005 to 2021 and build a predictive model of valve deterioration and reintervention risk at three, five and ten years.

TimeNet Challenge: An Online Hackathon for Time-Series AI

The TimeNet Challenge is an online hackathon that asks teams of 2–4 to onboard a time-series dataset in TimeNet format and train or fine-tune a Temporal AI Model such as OpenTSLM for tasks like forecasting, classification, anomaly detection or reasoning. Round 1 launched November 11 and closes November 22 at 11:59 p.m. PST, with more rounds to come. Nebius offers $100 or $500 vouchers per team per round, and winning teams receive prize credits starting at $25,000 from AWS and Google Cloud.

Show HN: Vending machine that takes words instead of money

Cultivated Corpora built a "Type to Pay" vending machine prototype that accepts freshly written human prose instead of money, dispensing snacks in exchange for text used as AI training data. The terminal is being demonstrated at the Data Vandals Newsstand at 51st and Lexington in Downtown 6 from Nov 12 through Nov 17. The company says it is filling the gap for AI firms that need more training data than currently exists.

US H-1B data: software developer filings fall 36% as AI-titled roles rise

US employer H-1B filings for software developers fell 36% from fiscal 2022 to fiscal 2026, comparing October through June, while filings with AI job titles rose and made up 1.8% of all certified filings in fiscal 2026. New-employment software filings fell 46%, and the fiscal 2026 drop was steepest from October to December 2025, down 26% year over year, after the White House announced a $100,000 fee on new H-1B petitions. Titles containing "machine learning" or "artificial intelligence" rose 80%, and shorter titles such as "AI engineer" supplied 92% of the latest year's increase; OpenAI and Anthropic included AI or machine learning in only 11% of their 402 certified filings.

Mirendil raises $1B Series A led by Kleiner Perkins to democratize AI R&D

Mirendil announced a $1B Series A led by Kleiner Perkins and Clearlake Capital, with participation from Battery Ventures, Google, M13, Pear VC and others, to build tools that make frontier AI R&D accessible to scientists and engineers. The company says one researcher working with agents pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months with less than $10M in compute, then post-trained it into Fern, which holds up to 10M tokens of context. Lor, now in private beta, lets hundreds of agents write code and run experiments for days at a time, and Mirendil says it runs on Lor every day.

Microsoft Research Focus Edition #98: AI Translation, KLS Conjecture Proof, and Faster Coding Agents

Microsoft Research Focus Edition #98 highlights AI advances bringing universal translation closer to reality and a new proof resolving the 31-year-old KLS conjecture in high-dimensional geometry. The edition also covers CompactionBench and APEC, which cut coding-agent context compaction costs by up to 25.4% and execution time by 9.6%, plus the Latent Core Tokenizer, which improved tokenization across 104 languages.

Google researcher's AI agents did not discover an unknown Maya site, archaeologists say

A Google researcher's team of 48 AI agents flagged a Maya site in Mexico as potentially unregistered, but archaeologists noted the site had already been documented in an academic journal more than a year earlier. Pietro Miotti, who described the work as a personal side project without Google compute or resources, said he checked with Mexico's INAH and later updated his post to credit the earlier work after criticism on social media. Archaeologists including Sarah Parcak and Ivan Šprajc said spotting a site is not the same as discovering it and stressed that peer review and existing research remain essential.

The AI 100: The 100 Private Companies Defining the Age of AI

Newcomer and Bessemer Venture Partners released the inaugural AI 100 ranking of the most promising private AI companies, with Anthropic at #1 ahead of its expected IPO next week and OpenAI at #2. The 100 companies carry a collective valuation of roughly $3.5 trillion, and the researcher-heavy model companies make up only 9% of the list but 55% of that valuation. Inference players Fireworks (#9), Baseten (#12), Fal (#16) and Modal (#20) also rank highly, alongside pivoted veterans like Stripe (#4) and Anduril (#7).

GDC Survey: 40% of Game Developers Now Use Generative AI on the Job

Forty percent of game developers say they use generative AI as part of their job, up from 36 percent in 2025, according to the GDC Festival of Gaming's latest annual industry survey. Usage is highest at free-to-play developers (58 percent) and mobile-only studios (70 percent), with research and brainstorming (73 percent) the top use case. The survey also found 34 percent of respondents were laid off sometime in the past two years.

What Huawei's Horinger data center reveals about China's compute gap

Epoch AI estimates Huawei's Horinger data center in Inner Mongolia holds about 105,000 operational Ascend 910C and 8,000 Ascend 910B chips, equal to roughly 81,000 H100-equivalents (90% CI: 36,000 to 141,000). The estimate comes from commercial satellite imagery and a cooling-equipment model that puts the four AI-focused buildings at up to 286 MW of IT power capacity. That leaves Horinger at about 7% of the AI compute in SpaceX's Colossus 2 in Tennessee, and a planned building could raise its capacity by over 70% by early 2028, still far behind Colossus 2's projected 1.8 million H100e.

ChinaTalk investigation: US data brokers sell about $500M in RL training environments to Chinese AI labs

US data brokers including Mercor, Surge AI, AfterQuery and Turing are selling roughly $500 million worth of training data, some of it RL environments, to Chinese AI labs, according to a ChinaTalk investigation. The report identifies at least nine US firms involved and says Chinese labs deliberately buy products already sold to Anthropic or OpenAI to skip costly in-house development, savings that matter given China's chip memory bottlenecks. ChinaTalk notes RL environments are largely unregulated by export controls and argues the Bureau of Industry and Security should apply existing controls to dual-use coding and biology environments, while stressing Chinese labs still trail US peers by at least a few months.

Plurai raises $13M seed to make AI agents more reliable via simulation

Plurai has raised $13 million in seed funding led by Team8 to make AI agents more reliable using simulation techniques borrowed from self-driving cars. The startup builds a digital twin of a company's systems, maps possible agent flows, and uses synthetic data to prepare models for edge cases, then post-trains small open-weight models tailored to each customer. Tests with Nvidia and Nebius showed its models outperforming midtier frontier models at lower cost.

Meta Signs Long-Term Purchase Agreements for Lower Carbon Cement, Steel, Trucking and Recycled Plastics

Meta has signed a set of long-term purchase commitments covering cement, steel, trucking and recycled plastics to bring clean manufacturing to commercial scale in its data center supply chains. The deals include at least 180,000 tons of lower carbon cement from Sublime Systems, near-zero-emission steel from Sweden-based Stegra, an expanded zero-emission freight route with Nevoya along I-10 in California and Arizona, and 5,000 tons per year of recycled PET with MacroCycle Technologies.

US data vendors are selling RL environments to Chinese AI labs, report finds nine vendors

US data vendors including Mercor, Surge AI, Turing, AfterQuery and Pareto are selling reinforcement learning environments to Chinese AI labs, and the report identifies nine vendors with China business. It argues these sales let Chinese labs skip costly rollout experimentation and save memory-constrained compute, while noting the deals fall largely outside export controls and that the lag behind US labs limits the damage.

Mercor raises $750M Series D at $20B valuation on $3B annualized revenue

Mercor raised a $750M Series D at a $20B valuation after reaching $3B in annualized revenue, up from $1B in February and $2B in June. The company reports 5 million experts and frames its market as the 800-plus US occupations whose workflows models have yet to be evaluated against. Mercor also says it already spends 3x as much on inference as on employee salaries, and argues that if the ratio holds across the economy it points to double-digit GDP growth over the next five years.

Ulanqab, a Remote Chinese City, Becomes an AI Boomtown and a Magnet for Young Workers

Ulanqab, a once little-known city in Inner Mongolia, has become an emerging AI computing hub as data centers spread across its grasslands, drawing young workers from across China. Kuaishou began operating its first self-built data center there in 2023, and DeepSeek advertised senior operations and maintenance roles paying up to 30,000 yuan ($4,200) a month this April. The labor market is now saturating, with employers raising hiring standards and some considering only bachelor's degree holders.

Ukraine's Fiber-Optic Drones Squeezed by AI Data Center Demand

Ukraine's fiber-optic drones, immune to battlefield jamming, now face higher costs as importing fiber material from China competes with global demand from AI data centers for the same limited supply. Ukrainian drone makers are adapting, with finished spools holding enough cable for drones to regularly operate at ranges of dozens of miles, and up to 62 miles in extreme cases.

AI-exposed workers took a month longer to find new jobs in 2025

Workers who lost jobs in AI-exposed roles in 2025 took about a month longer than peers in other roles to find new employment, according to Exponential View's data roundup. Anthropic paid $3.4 billion to take over a supplier's two-year-old order for 1.7GW of gas generators, the largest known power-equipment purchase by an AI lab. Apple's memory inventory grew 2.5x year on year, helping shield it from memory cost increases.

Lubrizol's AI-Driven R&D Strategy in Specialty Chemicals

Lubrizol is building a digital foundation that connects business processes, modern data platforms, and cloud infrastructure to speed up R&D for specialty chemicals. IT vice president Girish Rao says the aim is to give scientists better tools, not replace scientific judgment, with "fusion teams" of scientists, data scientists, and IT staff starting from small pilots to build trust. Cloud access also supports sustainability work such as modeling starch and corn-based material alternatives.

PFN's AI Writing Instruction System Ronjiro Wins AI Special Award (Learning Support) at 23rd Japan e-Learning Grand Prize

Preferred Networks' AI writing instruction system Ronjiro has won the AI Special Award (Learning Support) at the 23rd Japan e-Learning Grand Prize. Ronjiro uses generative AI including PFN's domestically developed PLaMo foundation model to support high school students' essay and composition learning, providing feedback on logical structure and improvement points in as little as one minute. It has been offered since March 2026 through YP Switch, a joint venture with the Yarakis Switch Group, and is now being introduced in some School IE classrooms nationwide.

SemiAnalysis visits Envision's 800MW Galaxy datacenter campus in Inner Mongolia

SemiAnalysis toured Envision's Galaxy campus in Ulanqab, Inner Mongolia, planned for up to 800MW of IT capacity, with its first 120MW phase delivered within a year at about a fifth of US costs. The campus runs a live 5MW 800V DC pilot using solid-state transformers that convert 10kV AC straight to 800V DC for racks, and drops the central chiller by sharing one water loop between liquid- and air-cooled servers. Envision supplies over 60% of the first two phases' power from its own 420MW wind farms plus 4-hour battery storage, and the full campus is expected to be completed by the end of 2027.

Apple Skips the AI Capex Race While Hyperscalers Spend 6x-9x More

Apple has become the only Mag7 company to beat a major index year-to-date, as its capex stays below pre-AI-spree levels while the four main hyperscalers raised spending roughly 6x-9x since 2023. Roblox said over 80% of its top 100 games by engagement used AI tools in the past month, with daily asset uploads tripling month-over-month.

I Watched a Tau Robotics Humanoid Robot Clean My Apartment

MIT Technology Review tested Tau Robotics' humanoid cleaning service in San Francisco, where robots are remotely piloted by humans rather than fully autonomous. The robots collect training data for future autonomous models, with pricing at $60 per weekly cleaning with two robots or $100 for a one-time visit. The robot cleaned a kitchen and toilet but struggled with stairs and repeatedly stalled during the first visit.

Amazon Scientist Hamed Karimi Wins 2026 ECML PKDD Test-of-Time Award

Amazon senior applied scientist Hamed Karimi and coauthors Julie Nutini and Mark Schmidt received the 2026 ECML PKDD Test-of-Time Award for their 2016 paper on linear convergence of gradient and proximal-gradient methods under the Polyak–Łojasiewicz condition. The paper showed that this 1963 inequality guarantees linear convergence even when models lack a single best solution, unifying scattered optimization results and extending them to nonsmooth landscapes and algorithms such as coordinate, sign, and stochastic gradient descent. It has been cited over 2,000 times and appears in graduate textbooks.

Ai2 and Google collaborate on NeuralGCM+ climate models

Ai2 has launched a climate modeling initiative backed by three-year funding from Google.org and technical support from Google Research to develop NeuralGCM+, an open-source model for faster, more reliable climate projections. The project extends Google Research's NeuralGCM, which combines physical equations for large-scale dynamics with AI for smaller-scale processes like cloud formation and rainfall, to simulate atmosphere and ocean together. Ai2 aims to cut regional temperature and rainfall errors over the past 50 years and release models and frameworks with documented workflows for outside researchers.

SemiAnalysis: SpaceX will become the world's largest AI cloud, with ~13GW targeted for 2027

SemiAnalysis reports that SpaceX now has 3.6GW of datacenter capacity under construction across multiple sites and has raised its 2027 estimate to roughly 13GW, up from the 10GW figure it gave on its first earnings call. The firm says it has already identified more than 9GW of feasible capacity and argues SpaceX may be the fastest datacenter builder in the world, citing Colossus 2 build times of 91 days for 210MW and a new Mississippi site targeted at about 4 months. It also flags a likely datacenter at the TeraFab chip fab site in Texas and a BESS deployment at Colossus 2 approaching 6GWh.

Sakana AI Symposium: Jürgen Schmidhuber on AI's Past and Future, Japan's Push into Language Models, World Models and Physical AI

Sakana AI held its "From World Models to Real-World AI" symposium in Tokyo on October 26, 2026, where Chief Scientific Advisor Jürgen Schmidhuber gave a keynote and Sakana AI announced it will develop its own pretrained language model series, Sakana Sawara (tentative name). The first small model, Sawara Small (tentative name), is planned for release in spring 2027 if development proceeds smoothly. More than 400 attendees joined the event, which also featured a conversation between Schmidhuber and CEO David Ha.

CXMT Pushes Into HBM3E, Unitree Q3 Revenue Up 70%, and China's Stock Slide

CXMT reported Q3 net profit of RMB54.4 billion and revenue of RMB118.71 billion, up 613.2% year on year, while Reuters says it has sent HBM3E samples to domestic chip designers with limited production planned within two months despite yields of just 20–30%. Unitree's first post-listing results showed Q3 revenue up 69.9% to RMB665 million as research spending rose roughly 250%. Moonshot launched Kimi K3.1 Pro with a one-million-token context window, then apologized for rate limits and restored a quarter of subscribers' monthly allowance.

Microsoft Research Focus Edition #97: ForkPilot, CABRA, TeleTune and AI Use at Work

Microsoft Research Focus Edition #97 highlights ForkPilot, a self-improving agentic framework that matches the best existing methods while cutting AI processing demands by up to 59% across six benchmarks. The edition also covers CABRA, a framework showing coding agents still struggle to understand unfamiliar codebases, and TeleTune, which learns reusable skills from user activity logs. It also examines Copilot Chat activity, where document editing exceeded 20% of classified activities.

Zenrin Group and Preferred Networks Form Business and Capital Alliance to Fuse Geospatial Data with AI

Zenrin, Zenrin DataCom and Preferred Networks have signed a business and capital alliance, with the Zenrin companies subscribing to a PFN third-party allotment. The partners will combine continuously updated geospatial data, including historical change records, with PFN's generative AI, physical AI and 3D spatial recognition technology to build AI models that understand and predict real-world change. Initial focus areas include public sector, real estate, retail and distribution, targeting demand forecasting, store site selection and urban planning, disaster prevention and logistics simulation.

Meta Supports AI Research at Five US Universities to Accelerate Medical and Scientific Breakthroughs

Meta is supporting research at five US universities — Caltech, Ohio State University College of Medicine, Louisiana State University, Indiana University, and UTHealth Houston — with access to its latest Muse models, researcher collaboration, and funding. Researchers will use Muse Spark Open, Muse Wonder, Muse Spark, the 30B open-weight Muse Glimmer, plus SAM, V-JEPA, DINO, UMA, and SeamlessM4T. Projects span medicine, physics, chemistry, biology, computer science, and mathematics, with universities retaining control of all patient and clinical data.

Anthropic reports early progress on using Claude to accelerate drug development

Anthropic's life sciences team has spent several months working with six drug developers, including Bristol Myers Squibb, Generate Biomedicines and Divergence Labs, plus running its own discovery programs, to test where Claude can speed up drug development. Reported early results include a Generate design roughly 70% shorter than the standard version that let immune cells make about 10 times as many receptors, and roughly 35,000 antibody designs generated largely autonomously for Divergence in about a month. Anthropic says the work is still mostly preclinical and plans to extend Claude into clinical trials as the programs advance.

Anthropic Opens Call for Guest Investigators in the Life Sciences

Anthropic is inviting principal investigators and their research groups at academic institutions, nonprofits, and government labs to apply for 8–12 week guest investigator projects using Claude in areas such as causal human biology, structural and computational molecular biology, medicinal chemistry, drug discovery, and clinical science. Anthropic will run its models and harnesses at no cost, and outputs will be published with open-source data and code while investigators retain rights to their data. Applications are open for approximately 14 days, with selected participants notified by email about 14 days after the close.

Epoch Brief, October 8, 2026: AI Agent Population Estimates, Falling AI Costs, and China Chip Exposure

Epoch AI estimates that AI chips shipped through 2027 could support about 30 to 170 million concurrent frontier-model agents, or almost 2 billion if run on cheaper models. Its research also finds the cost of a fixed level of AI performance has fallen about 47% per quarter over three years, roughly 13× per year. Separately, Epoch estimates China's exposure to semiconductor supply shocks is 2.7× that of the US.

Markov Robotics Uses Modal to Run End-to-End Robot Training

Markov Robotics is training a world model for zero-shot general-purpose robot dexterity, using fewer than ten hours of robot data where comparable results typically need thousands of hours. The company runs nearly all its compute on Modal, from preprocessing hundreds of terabytes of video to 2–10 concurrent training runs and remote inference via Modal Tunnels at roughly 20 milliseconds round-trip latency. Preprocessing that could take a full day completes in twenty to thirty minutes, with fan-out to upwards of 500 GPUs at once.

Applied Compute scales reinforcement learning for custom enterprise agents on Modal

Applied Compute trains custom enterprise AI agents via reinforcement learning, building on Modal's sandboxes, Functions and sub-second cold starts to keep thousands of parallel rollouts GPU-bound. Its agents include a merchant onboarding model for DoorDash and a bug-catching agent for Cognition.

Chai Discovery scales computational drug discovery on Modal

Chai Discovery is using Modal to run its machine learning drug discovery pipelines, from protein embeddings and multiple sequence alignments to antibody design. Modal Volumes downloads and indexes datasets reaching hundreds of gigabytes once, then shares them across machines, while elastic GPU scaling spins up hundreds of GPUs in minutes. Chai says the same setup now carries research workflows into production without rewriting infrastructure.

Twirl Joins Modal to Expand Data Orchestration and European Presence

Modal announced that Twirl, a data orchestrator for developing, testing and deploying data pipelines, is joining the company. Twirl co-founders Rebecka Storm and Eric Hansander will join Modal's Stockholm office, where they will develop and expand Modal's serverless infrastructure and grow its European footprint. Modal said the team's data product experience will help it expand beyond AI inference into adjacent AI, ML and data use cases.

How a Top-Tier European Soccer Team Cut Data Processing Costs by 50% with Modal

A top-tier European soccer team, anonymized as AFC Richmond, moved its match tracking data pipeline to Modal's serverless GPU platform and cut the cost of processing a full season of games by 50%. The team runs inference on ~3.5 million tracking observations per game through a custom transformer-based model, and with Airflow on Modal now processes each game's data in minutes rather than hours.