NAll news

All news

Monday389 items

  1. 36Kr · AI
    AI score65

    Meta and Fourier Intelligence Partner to Build Humanoid Robot Prototype Max in Three Months

    According to foreign media reports, Meta partnered with Chinese robotics company Fourier Intelligence to develop a humanoid robot prototype named Max in about three months, and a second-generation product has already entered the R&D stage. Max is about 1.5 meters tall and weighs 48 kilograms, is equipped with NVIDIA's Jetson Thor chip, has five-fingered dexterous hands and a display screen that can simulate human facial expressions, and needs to recognize human voices and gestures and maintain eye contact. Li-Chen Miller, Meta's Vice President of Robotics Product, was stationed in Shanghai for nearly two months to help advance the effort, and Meta is also recruiting for supplier development and mass-production preparation roles in Shanghai and Shenzhen.

    Same story · 6 reports

    #Industry news#Embodied AI#Meta

  2. 36Kr · AIOpinion
    AI score55

    Three Goldman Sachs Veterans Judge That the Peak Growth Rate of AI Capital Expenditure Is Near, and the Theme Is Shifting to a Conventional Narrative

    In a podcast recorded on October 29, Tony Pasquariello, Goldman Sachs' global head of hedge fund coverage, Josh Schiffrin, global head of risk, and Dominic Wilson, senior advisor at Goldman Sachs Research, judged that the peak growth rate of AI capital expenditure is already quite close, and that AI is turning from a super-theme that has lasted about three years into a more conventional narrative. Wilson listed three cracks—pressure in credit spreads, doubts about the monetization path, and earnings divergence—but stressed that he was talking about a peak in the growth rate, not an absolute decline in scale. Schiffrin believes the market has priced in 60 to 65 basis points of further rate hikes, that the actual number may be smaller, and listed going long global front-end rates as his highest-conviction trade for year-end.

    #Industry news#OpenAI#English text is being prepared.

  3. 36Kr · AIOpinion
    AI score71

    Embodied intelligence funding cools: September funding down 40% month-on-month, old shares hard to sell even at a discount

    Tencent Technology reports that in September 2026, 46 financing deals occurred in China's embodied intelligence primary market, with a reference amount of approximately 11.624 billion yuan, down 4.2% and 40.1% respectively from August. Multiple investors said new project valuations are doubling too fast, and old-share transfer quotes have fallen to 65% to 80% of the previous round's valuation. After initiating preparations for a Hong Kong listing, Zhiyuan shifted to advancing a new financing of 1 billion to 1.5 billion yuan at a valuation of about 50 billion yuan. The article also mentions that Unitree's stock price has retreated from its first-day high after listing, and that the IPO processes of some robotics companies are stalled and their revenue quality is being questioned.

    #English text is being prepared.#Embodied AI#Industry news

  4. 36Kr · AIResearch
    AI score58

    LeCun and Balestriero Publish JEPA Paper: 109-Million-Parameter Model Surpasses… in Cross-Modal Representation Alignment

    Yann LeCun and Brown University assistant professor Randall Balestriero have published a 37-page paper, "Proving and Improving the Platonic Representation Hypothesis with JEPAs," proposing that under positive-sample relationships meeting specific conditions, JEPA models of different modalities can learn representation spaces that differ only by a rotation transformation. In experiments, a 109-million-parameter text JEPA scored 0.184 on the cross-modal representation alignment metric, higher than Qwen2.5-72B's 0.154 and Llama-3-70B's 0.151, but this metric measures the closeness of representation structures, not general capability. The research also claims that Good Views training reaches the alignment level of next-token prediction with 900 million samples using about 21.6 million samples, a gap of roughly 14 times when calculated by token count.

    Same story · 3 reports

    #Papers/Research#English text is being prepared.#Multimodal

  5. 36Kr · AI
    AI score36

    AI Cannot Cure the CEO's Serious Illness: Capital Begins to Price in "AI Risk"

    In October, the 80% correlation between crude oil and US Treasury yields that had held for a year suddenly broke down. BlackRock portfolio manager Jeffrey Rosenberg believes geopolitical risk has given way to AI risk, with the Mag 7 collectively pulling back and Google's market value evaporating by nearly 500 billion in a single week. Santander's AI projects created 194 million euros in business value in the first three quarters, while Microsoft issued 25 billion dollars in bonds to double down on AI and its CDS spread continued to climb, as the market began to price AI in tiers. The article points out that most "AI autonomous agents" are actually backed by human outsourcing, and during internal testing of Meta's Muse agent, a large number of calls were handled by real people at call centers.

    #English text is being prepared.#Agents#Industry news

  6. 36Kr · AIOpinion
    AI score74

    Anthropic's rating pressure and the cash flow strain of Big Tech's AI capital expenditure

    Blackstone expects that Anthropic may remain at the BB level even after completing an IPO, and S&P changed its rating outlook for Meta to negative on the same day. In the first three quarters of this year, the capital expenditure of Microsoft, Alphabet, Amazon, Meta and Oracle totaled $556.9 billion, exceeding operating cash flow for the first time, and the combined free cash flow of the five companies turned from positive to negative. The article also outlines the structure through which institutions such as Blackstone arranged $63 billion in financing for Anthropic's computing power procurement via special purpose entities.

    #Industry news#English text is being prepared.#English text is being prepared.

  7. TMTPost · AGI
    AI score41

    Alibaba Cloud's You Liang Discusses Hyper-Convergence of HPC and AI: How E-HPC Unifies Scheduling of HPC, AI, and Agents

    You Liang, head of high-performance computing at Alibaba Cloud, proposed that the key to hyper-convergence of HPC and AI is to use a unified platform to simultaneously schedule HPC, AI training and inference, and agent tasks, and to distill industry expert experience into Skills through the E-HPC agent platform. Alibaba Cloud says that more than 80% of time in the scientific research process still relies on manual handoffs, with actual compute execution accounting for less than 20%; its hpc9a instances are equipped with fifth-generation AMD EPYC processors with up to 128 cores, delivering about 30% higher performance than the previous generation. E-HPC has developed dedicated solutions for chip design, AI4S and life sciences, and embodied intelligence, among which HPDA Robotics can improve the cost-effectiveness of multimodal data processing by 2.1 to 5 times.

    #Industry news#Agents#English text is being prepared.

  8. TMTPost · AGIOpinion
    AI score60

    Kimi Apologizes Three Times in Six Months: The Stronger the Model, the Harder the Business?

    A TMTPost analysis points out that less than 12 hours after Kimi K3.1 Pro launched, it sparked controversy due to rapid quota consumption; Moonshot AI subsequently apologized and restored 25% of the monthly quota in a one-time credit to all plan subscribers, while also stating it would reduce unnecessary sub-Agent calls and output length. The article says this is Kimi's third apology within six months, and citing Artificial Analysis data, notes that K3.1 Pro produced a cumulative output of about 150 million tokens on the Intelligence Index, roughly 72% higher than the median of the models tested. The article also mentions that Moonshot AI's compute relies on external supply, and compares the different monetization paths of vendors such as DeepSeek, Zhipu, and MiniMax.

    #Industry news#Agents#DeepSeek

  9. TMTPost · AGIOpinion
    AI score71

    Blackstone lowers Anthropic rating expectations; massive AI capital expenditure sparks credit concerns

    Blackstone expects that Anthropic may remain at the BB junk-grade level even after completing an IPO; previously it had expected the company to obtain an investment-grade rating by the end of this year. In the first three quarters of this year, the combined capital expenditure of Microsoft, Alphabet, Amazon, Meta and Oracle reached $556.9 billion, exceeding operating cash flow for the first time, with their combined free cash flow turning from positive to negative. S&P changed its rating outlook on Meta to negative, Oracle was downgraded to BBB- in July, and Blackstone is still working with Bank of America, Citi and Morgan Stanley to advance a $63 billion Anthropic computing power financing deal.

    #Industry news#Anthropic#OpenAI

  10. TMTPost · AGI
    AI score40

    AI Can't Cure the CEO's Serious Illness: Capital Begins to Price in "AI Risk"

    Jeffrey Rosenberg, a portfolio manager at BlackRock, pointed out that in October the 80% correlation between crude oil prices and U.S. Treasury yields that had held for a year suddenly broke down. In the market's eyes, geopolitical risk has given way to AI risk, with the Mag 7 collectively pulling back and Google's market value evaporating by nearly 500 billion in a single week. Santander's AI projects created 194 million euros in business value in the first three quarters, while hyperscale cloud providers' capital expenditure has outpaced free cash flow, and Microsoft issued $25 billion in bonds to double down on AI. The "AI agents" of companies such as Dazzle AI and Duckbill still rely on human outsourcing as a backstop in real-world deployment, and Meta's internal testing of Muse also relies heavily on real people at call centers standing in.

    #English text is being prepared.#Agents#Industry news

  11. heise online
    AI score46

    Meta Reportedly Collaborating with China's Fourier to Develop Humanoid Home Robot Max

    Meta Platforms has reportedly partnered with Chinese robotics startup Fourier to develop a humanoid robot prototype named Max for everyday household chores, but it is an internal research project and will not be sold externally. Max weighs approximately 48 kilograms and stands about 1.5 meters tall, is based on the NVIDIA Jetson AGX Thor platform released in August 2025, and the two parties built the first prototype in about three months, with development of Max 2 already underway in September.

    Same story · 6 reports

    #Industry news#Embodied AI#OpenAI

  12. heise online
    AI score24

    BSI Warns German Federal Politicians That Their iPhones May Be Targeted by Spyware; Meta Teams Up with Fourier to Develop Humanoid Robots

    The German Federal Office for Information Security (BSI) has warned members of the Bundestag and government officials that their iPhones may be targeted by Mercenary spyware, and Apple has been issuing intrusion warnings to affected devices since October 29. Meta has partnered with Chinese robotics startup Fourier to develop a humanoid robot prototype named Max, which is about 1.5 meters tall and weighs 48 kilograms, to train its own AI models to perform household tasks, and it does not plan to sell it externally. OpenAI CEO Sam Altman, meanwhile, called running AI models locally a waste of resources, arguing that DRAM in private hands is "only used 5% of the time."

    #Industry news#Embodied AI#Safety / alignment

  13. English text is being prepared.
    AI score44

    Samsung's first AI glasses set to launch in South Korea and the US in mid-December

    Samsung Electronics has set the launch date for its first AI glasses in mid-December, with South Korea and the US as the first sales markets; pricing has not yet been finalized. This initial launch does not cover major markets such as Europe, China and Japan, and the sales scope is noticeably smaller than that of the Galaxy S26 series, which covers about 120 countries and regions. Samsung previously also launched its headset Galaxy XR first in South Korea and the US, only expanding to the UK in July this year.

    #Industry news#English text is being prepared.

  14. English text is being prepared.
    AI score18

    Varjo announces that its XR-4 series headsets will fully shift to domestic production in Finland, with assembly to begin in January 2027

    Starting in January 2027, Varjo will assemble its XR-4 series headsets at its factory in Espoo, Finland, expanding domestic manufacturing from the existing XR-4 Secure Edition to the entire product line. The company says the move is intended to reduce the risk of supply chain disruptions, and that all headsets it manufactures in Finland will comply with the requirements of the U.S. Trade Agreements Act (TAA). Varjo technology is currently used in more than 120 defense projects across NATO and allied countries.

    #Industry news#English text is being prepared.

  15. Semiconductor Industry Observer
    AI score46

    Samsung: Memory Demand Squeezes Foundry Capacity, Sub-8nm Process Utilization to Exceed 100% by 2027

    Samsung Electronics expects that by 2027 demand for all sub-8nm processes will exceed capacity, with capacity utilization surpassing 100%, and 4nm revenue will continue to rise on the back of AI and HPC orders from a leading US customer and steady memory demand. Samsung Foundry is producing 4nm logic base dies for its own HBM4, has secured more 2nm projects in the third quarter, and the number of 2nm projects in 2026 is expected to more than double compared with last year. New 2nm products for leading customers will not enter mass production until the second half of 2027, and 2nm yield is still stabilizing.

    #Industry news#English text is being prepared.

  16. Semiconductor Industry Observer
    AI score47

    Google: Agentic AI Drives Up Data Center CPU Demand

    Mark Lohmeyer, Vice President of AI and Compute Infrastructure at Google Cloud, said that the bursty workloads and long-duration state retention of agentic AI are driving up the share of data center CPU compute, with CPUs rising from a previously small share to a level that cannot be ignored. Google Cloud uses microVMs for agent isolation, increasing the number of virtual machines that can be hosted on the same hardware from 100 to 1,000 microVMs, and simultaneously offers Intel and AMD x86 CPUs along with its self-developed Arm-based Axion CPU for customers to choose based on workload.

    #Industry news#Agents#English text is being prepared.

  17. Semiconductor Industry Observer
    AI score48

    NVIDIA to stop supplying RTX 5090 DV2 and RTX 5080 16GB; China's flagship gaming cards will retain only…

    NVIDIA will stop supplying the RTX 5090 DV2 and RTX 5080 16GB, leaving only one flagship gaming card in the Chinese market, the RTX 5080 24G, which will launch as early as December this year or as late as January next year. The new product is expected only to expand VRAM from 16GB to 24GB, a 50% increase, with core specifications unchanged; the current street price of the RTX 5080 16GB in stock generally exceeds 20,000 yuan, and the supply chain judges that the 24GB version will most likely be priced above 30,000 yuan.

    #Industry news#English text is being prepared.

  18. Analytics India MagazineProduct release
    AI score52

    Arrowhead AI Releases Fletch Speech Model, Supporting 14 Languages and Used for Over a Million Customer Service Calls Daily

    Arrowhead AI, a Bangalore, India-based speech AI startup, released its self-developed TTS model Fletch on October 30, targeting multilingual customer calls in India and Southeast Asia. It has been deployed in real customer service scenarios, supporting more than one million calls per day. Fletch supports languages including English, Hindi, Tamil, Telugu, Bengali, Malay, and Indonesian, can switch languages within the same sentence, and offers 12 emotional styles such as empathetic, apologetic, calm, and firm. The company says its latency is 80 milliseconds, with speech generation beginning in under 100 milliseconds, and that pronunciation accuracy improves by about 30% on selected words such as brand names, product names, customer names, and Indian proper nouns. It is currently used for collections, sales, renewals, and inbound customer service, with customers including Axis Bank, RBL Bank, IndusInd Bank, Tata 1mg, upGrad, KreditBee, Kissht, and TVS Credit.

    #Model releases#Speech#Multimodal

  19. Unite.AIResearch
    AI score67

    SemiAnalysis releases the GDPx benchmark, covering 11 agent domains

    SemiAnalysis releases the GDPx benchmark, covering 11 professional domains including chip design, cybersecurity, biology, finance, and law. The first version contains 206 tasks built by 14 data partners, and the leaderboard includes 13 models. Claude Opus 5.5 ranks first at 12.0%, with an average cost of $51.41 per task and 548K output tokens; Gemini 4 Carbon ranks second at 10.9%, and GPT-6 Astra ranks third at 9.6%. SemiAnalysis says the benchmark is updated every 3 to 6 months, and criticizes OpenAI GDPval's tasks as non-interactive, with context written directly into the prompt.

    #Evaluations / benchmarks#Agents#English text is being prepared.

  20. gihyo.jp
    AI score66

    OpenAI, Google, and Microsoft will support the __usage_ai-trainin… identifier in robots.txt

    On October 30, 2026, OpenAI, Google, and Microsoft announced that they will support using the __usage_ai-training identifier in robots.txt to uniformly express to multiple AI vendors whether content may be used for training. This mechanism is based on the draft specification Machine-Readable Training Preference Signal, published after discussions in the IETF AIPREF, and was submitted by Anthropic's Kevin Kelley; it is not an IETF-approved standard. The three have different implementation timelines: Google plans to support it in early 2027, Microsoft will apply it starting February 1, 2027, and OpenAI says it will implement it within the coming months; Anthropic has also begun implementation.

    Same story · 2 reports

    #English text is being prepared.#English text is being prepared.#OpenAI