Create

Sign in to ReadmeX

or
ReadmeX
ReadmeX

A clearer picture, in a conversation.

Catch up on what matters, then ask a little deeper.

Your community and people briefings stay personal to you.

AI News

‹Thu, October 8›
All dates
2014
Industry

CCTV exposé: AI digital-human 'doctors' push fake medical ads for ¥48 a week

A CCTV "Weekly Quality Report" investigation exposed AI digital humans posing as medical professionals to endorse drugs and health supplements, with generation costs as low as 48 yuan per week — one app turns a single photo into an 8-second digital-human ad in minutes and, according to the article, applies no checks on claimed professional identity. The piece compares tooling pricing and features, including HeyGen's $29/month Creator plan and Alibaba Cloud Model Studio's pay-as-you-go rates (0.5 yuan/sec for 480P video, 0.9 yuan/sec for 720P), and notes that combinations such as "medical worker + drug" are generally not blocked upfront. Enforcement cases are emerging: Foshan Xinyu Technology was penalized for a 65-second AI-synthesized "Wen Tiejun" product pitch ruled false advertising, and a Beijing company was fined 5,000 yuan for using an AI virtual human in a livestream to claim a ordinary food boosts immunity — a case the market regulator called the first of its kind after the February 2026 livestream e-commerce rules.

36氪 人工智能·
2024
Industry

From 8-GPU servers to supernodes: AI data centers for the inference era

Speaking at the 2026 ITValue Summit on September 17, xFusion computing vice president Wang Liangdong said AI has entered an inference-dominated phase in which global token consumption keeps surging and buyers judge infrastructure by cost per token, throughput and cluster stability rather than peak per-card performance. The article cites a range of figures: the 2026 China Development Forum said China's daily token calls topped 140 trillion in March, OpenRouter data showed China accounted for 36% of global token calls in the March 16-22 week, JPMorgan expects Chinese inference token consumption to grow roughly 370x in five years, and IDC projects annual global consumption of 150,000 Peta Tokens by 2030. It argues the 8-GPU server paradigm hits communication bottlenecks under trillion-parameter, high-concurrency inference, pointing to Huawei's Ascend 960 supernode (4,096 cards per node, 8E FP8), Lenovo's Wanquan heterogenous computing platform V5.0 and xFusion's FusionPoD for AI rack-scale designs — while warning that supernodes only suit very large workloads, interconnect standards remain unsettled, and utilization depends on sustained paying demand.

钛媒体·
2036
Products

Google Maps' Ask Maps may soon let you share conversation links, and could be renamed Ask

A teardown of Google Maps version 26.41.00 by Android Authority surfaced a "Public link to conversation" string, suggesting Ask Maps could let users share entire chats via public links much like the Gemini app. Today users can only copy individual Ask Maps answers, and the copied text drops the URLs of places mentioned. The teardown also hints the Gemini-powered feature may be shortened to Ask and moved into the app's search bar, though the author could not actually trigger the public-link option and cautions it may never ship.

Android Authority·
2046
Industry

Kyndryl survey: only 10% of IT chiefs use agentic AI for legacy modernization

A Kyndryl survey of 2,000 senior IT decision-makers found that just 10% of technology chiefs apply agentic AI to modernization work, four in ten enterprises still run mainframes, and 18% say their modernization efforts delivered limited or unclear value. Andy Thurai, founder of The Field CTO and a former IBM chief strategist, said vendors pitch AI to IT leaders as a cure-all but "the economics don't hold up" — and three in ten respondents already worry AI-driven modernization will bring wildly unpredictable compute costs.

ZDNET AI·
2054
Industry

AI study camps for kids: ¥2,980 for a Big Tech tour that turns into a hard sell

Since this summer, orders for “AI study camps” in China have grown 370% year on year, with prices ranging from a few thousand yuan for a three-to-five-day program to tens of thousands, and destinations shifting from Peking and Tsinghua to ByteDance and Alibaba campuses. Parents say the Big Tech portion amounted to walking an open area and eating in the canteen, after which children and parents were split up and the adults attended a “wise parents” session that ended in a pitch for a multi-year, tens-of-thousands-of-yuan “focus” course; refund requests were refused on the grounds that the itinerary had finished. Educators and parents quoted in the piece say the camps' real value today is mainly tied to competitions and certificates used to pad school applications.

虎嗅 AI·
2064
Industry

US VC deal value hits record $515.8B as AI takes 82.7% share

US venture capital deal value reached $515.8 billion in the first nine months of the year, about 44% above the previous annual record, with AI accounting for a record 82.7% of that value, according to the PitchBook-NVCA Venture Monitor. Third-quarter deal value fell about 40% to $98.4 billion, and exits remain weak: SpaceX's $60 billion all-stock purchase of Cursor developer Anysphere alone made up 53.1% of quarterly exit value, leaving just $53 billion without it. PitchBook counts 179 new unicorns this year and 992 companies valued at $1 billion or more, and its model gives Anthropic an 86% chance of an IPO within a year, versus 12% for OpenAI.

SiliconANGLE AI·
2076
Industry

VEIR raises $110M to swap copper for superconducting cables in AI data centres

Massachusetts-based VEIR has closed a $110 million Series C, co-led by Matter Venture Partners and Tyche Partners, which the company described as oversubscribed. LG Technology Ventures, Gates Frontier, Sabanci Climate Ventures and Hui Capital joined as new investors, with existing backers including Engine Ventures, Galvanize Climate Solutions, Piva Capital, Congruent Ventures and VXI Capital also participating, bringing total funding to $225 million. VEIR pairs superconducting cables with cryogenic cooling, controls, terminations and monitoring, claiming it moves more power through less space; its first focus is behind-the-meter power at AI data centres, where it has demonstrated a 3-megawatt system but has not named customers.

The Next Web·
2086
Industry

LegalOn halves Codex costs by tiering GPT-6 Astra, Sol and Luna

After embedding Codex in its development process, LegalOn Technologies initially gave engineers unlimited access to GPT-5.5 in Fast mode, which pushed up spending. Its AI-powered Development CoE then introduced model-selection guidelines: GPT-6 Luna for everyday implementation, GPT-6.1 Sol for standard design, data analysis and documents, and GPT-6 Astra for complex analysis, architecture and agent orchestration, alongside restricting Fast mode by default and setting monthly budget caps per department, group and individual. OpenAI says the company cut estimated daily costs by roughly 65% versus GPT-5.5 while maintaining development speed, with mature businesses asked for about 20% better cost efficiency and new businesses given generous budgets.

OpenAI·
2096
Industry

Pollo AI's video agent runs on GPT-6 Astra and GPT-Image-2.5

OpenAI published a customer story on Pollo AI, whose Pollo Agent uses OpenAI models to turn rough creative ideas and visual references into storylines, scene flows, scripts and finished images and videos; Pollo AI says it serves more than 26 million users and that 30% of creator sessions start from video templates or trending-format workflows. According to the piece, Pollo AI uses GPT-5.6 for routing and GPT-6 Astra for harder work such as narrative development and scene revisions, while GPT-Image-2.5 handles image generation for posters, storyboards and campaign visuals. Its Photo to Video Ads tool is described as turning a still product photo plus a prompt into a roughly 15-second video ad. The user count, time-savings and quality claims are vendor-reported and not independently verified.

OpenAI·
2104
Research

Weinan E: AI for Science's second half is rebuilding the research system

Speaking at PAIR Conference 2026 in Shanghai, Chinese Academy of Sciences academician Weinan E argued that AI for Science has two layers: as research becomes “industrialized,” it needs new infrastructure — compute, models/algorithms/software, databases and knowledge bases, and automated labs (China has pushed this “platform research” approach since 2022) — while the “second half” is change in the research system itself. He said theory and computation matter alongside automated labs and literature mining: DPA-4 now covers organic molecules, catalysis and complex interfaces, but is still two to three orders of magnitude from materials and process design, and a “science foundation model” built by fine-tuning open general models may be a false premise if the underlying base model is weak. He also called for breaking three boundaries: theory/experiment, between disciplines, and research/industry.

MIT科技评论中文·
2114
Products

Microsoft launches MAI Code1.1Flash as GitHub Copilot adds local-cloud hybrid inference

At its October 7 Windows and Surface event, Microsoft said GitHub Copilot will gain support for local AI models by the end of this month, letting developers switch between cloud and on-device models or have the system schedule them automatically. The accompanying MAI Code1.1Flash is a mixture-of-experts model that Microsoft says has 137 billion total parameters and 6.8 billion active parameters, using quantization and speculative decoding to cut memory use and speed up on-device coding responses; it debuts on the new Surface Laptop Ultra with NVIDIA RTX Spark hardware. Users can pick inference modes through the GitHub Copilot CLI, the Copilot app and VS Code, or call local models via a Windows ML provider or an OpenAI-compatible local endpoint, with code isolation handled by Microsoft's Execution Container (MXC).

AIbase AI新闻·
2124
Open source

Tencent open-sources EVIE-4.5B visual document retriever

Tencent Hunyuan released EVIE-4.5B on Hugging Face under Apache-2.0, a visual document retrieval model built on a ColQwen3.5 late-interaction multi-vector architecture with a single-projection Prefix-MRL head that can be truncated at runtime from 2048 down to 64–2048 dimensions without separate checkpoints. The model card reports 66.02 on ViDoRe V3 (66.75 for the 8B flagship teacher) and pairs it with training-free HAC token compression that cuts roughly 750 vectors per page to 32, bringing a 1M-page index to about 3.81 GiB. Tencent says weights, training pipelines (including ARD distillation from the 8B teacher), the HAC algorithm and evaluation suites are all open-sourced, with a formal paper to follow.

腾讯混元 · Hugging Face·
2136
Industry

The Verge reviews Artificial, Luca Guadagnino's biting Sam Altman biopic

Artificial, Luca Guadagnino's satirical Sam Altman biopic, premiered at the New York Film Festival with Andrew Garfield as Altman, Yura Borisov as Ilya Sutskever and Ike Barinholtz as Elon Musk. The Verge senior AI reporter Hayden Field writes that the film tracks real milestones — Sutskever's time with Geoffrey Hinton, the 2017 Dota 2 win, Altman's ouster and return — and leans lightly on invented characters because lawsuits and leaked communications supplied the material, while still being darkly funny. The piece also notes a pre-screening remark thanking Neon for "standing behind this film," implying Amazon MGM Studios dropped it in June, presented alongside a mention of Amazon's $50 billion OpenAI investment as statements made at the event.

The Verge AI·
2144
Industry

Hunan bets on industrial and embodied AI with 10-billion-yuan fund

Hunan is focusing its AI push on industrial AI, robotics and embodied intelligence rather than head-on competition in foundation models. Officials say the province's intelligent-robot industry reached nearly 50 billion yuan in scale in 2025 with more than 300 companies, and it targets 100 billion yuan in industry-chain revenue by 2028 and 200 billion yuan by 2030. The Jinfurong Dachen Future Industry Fund, based in Xiangjiang New Area, has a total size of 10 billion yuan with a 3 billion yuan first phase, targeting early- and growth-stage AI, life-science, quantum and advanced-materials companies.

虎嗅 AI·
2156
Industry

Voltus raises $225M Series D to turn homes and batteries into data centre power

Voltus, a San Francisco company that runs a virtual power plant, raised a $225 million Series D led by Generation Investment Management, with Activate Capital and energy trader Vitol as co-leads and Broadscale Group, Climate Investment, NGP, Ajax Strategies and All Aboard Fund also participating. Voltus pays businesses and homes to cut, shift or supply power when the grid needs it, then sells that flexible capacity into electricity markets; it says its platform holds 8.5 GW today and targets 20 GW by 2030 across the US and Canada, and it did not disclose a valuation. The money will expand its Bring Your Own Capacity programme, which includes a three-year June deal with Google for up to 100 MW a year in PJM and an August agreement with Sunrun to supply home solar and battery capacity for Voltus deals with AI hyperscalers.

The Next Web·
2164
Industry

Yiwei Aerospace raises over RMB100M to build orbital AI compute for LLM inference

Yiwei Aerospace, a Chinese startup founded in February 2025, is pursuing 'space computing' — placing high-performance AI payloads on satellites to convert orbital energy into tokens. It says it has completed four funding rounds since July 2025, the latest an angel series of more than RMB100 million, with backers including the Beijing Aerospace Fund, Haohan Data and Century Golden Resources. Its lineup spans the Jiyu satellite components, the Dengyun space compute cluster aimed at large-model inference, the Tianyan payload SDK positioned against CUDA, and a space compute service platform; the team claims an 18% efficiency gain verified at the heat-dissipation stage.

品玩 实时要闻·
2174
Industry

Stripe: AI token fraud surges as hybrid pricing takes hold

Speaking at its Stripe Tour event in Shanghai, Stripe said more than one in six signups at AI companies involve multi-account abuse, with free-trial abuse and usage-based "dine-and-dash" billing also rising. Its Radar fraud product blocked over 3.3 million high-risk free-trial attempts in 30 days across eight businesses that deployed it, the company said. Stripe also reported that the share of AI firms using hybrid pricing rose from 27% to 41% over the past year, while fixed-subscription or per-seat pricing among top AI companies fell from 40% to 24%. It said its acquisition of OpenRouter is still pending and introduced Checkout Studio, a Machine Payments Protocol, and Stripe Billing with token-based metering.

品玩 实时要闻·
2184
Products

Tencent's Knocket turns personal pages into chat-enabled AI business cards

Tencent RTC has launched Knocket, whose centerpiece is a Contact Page: users build a standalone page with avatar, bio, portfolio, services, pricing, case studies, FAQ and booking links, and visitors can chat right on the page instead of hunting for a WeChat or email address. The page can also be wired to an AI agent and knowledge base that answers common questions from course or product material, handing complex queries to the human, and Knocket can generate the page with AI. The article says many core features are free and users can plug in their own AI models, framing Knocket less as AI customer service than as a single entry point for a personal business.

36氪 人工智能·
2194
Industry

AI is becoming the front door to healthcare, from Doubao to Ant AQ to ChatGPT Health

The piece traces AI's shift into the first stop for health questions: by March 2026 China's AI-native apps reached 446 million monthly users, with Doubao at 345 million, while Ant launched standalone AI health app AQ in June 2025 and by 2026 reported more than 100 million users and over 10 million health consultations a day. OpenAI said more than 230 million people a week asked ChatGPT health questions when ChatGPT Health launched in January 2026, rising past 300 million by July. The article stresses that AI handles pre-visit questioning and record organisation while diagnosis and treatment advice stay with doctors — a line both JD Health and OpenAI explicitly keep.

36氪 人工智能·
2204
Industry

OpenRouter data: OpenAI and Anthropic split business spending nearly evenly in September

The Wall Street Journal, citing OpenRouter data, reports that business spending on OpenAI and Anthropic models was roughly even in September, compared with Anthropic holding a 75% share in January. The report says both companies need to show Wall Street they have sustainable businesses ahead of planned IPOs.

Techmeme·
Next →← Previous