NAll news

All news

Wednesday626 items

  1. MIT News · Artificial Intelligence
    AI score28

    MIT Professor Paris Smaragdis Builds Machine Hearing Systems That Can Listen and Discern

    The research group of MIT Professor Paris Smaragdis is working to give machines hearing capabilities that surpass those of humans, for use in scenarios such as speech recognition, music understanding, and environmental monitoring. Their research covers machine learning for sound source separation, intelligent devices that can enhance human auditory environments, and tools that help people understand, manipulate, analyze, and create music. Smaragdis joined MIT in 2025 and holds appointments in the Music and Theater Arts program and the Department of Electrical Engineering and Computer Science.

    #Industry news#Speech

  2. East Money
    AI score60

    Anthropic's prospectus discloses mixed use of four types of chips, potentially weakening Nvidia's pricing power

    In the S-1 registration statement published on Monday, Anthropic said it has integrated computing resources from multiple strategic partners into a unified, flexibly allocatable pool, and allocates tasks according to each chip's performance advantages in different scenarios. It uses four types of chips to train Claude and provide services: Amazon Trainium, Google TPU, Nvidia GPU, and AMD GPU. Patrick Moorhead of Moor Insights & Strategy believes that customers able to shift tasks across four platforms are in a different position in negotiations, and what may really be affected is Nvidia's pricing power; IDC analyst Ashish Nadkarni said a multi-chip strategy can also reduce sensitivity to fluctuations in Nvidia chip supply.

    Same story · 124 reports

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

  3. East Money
    AI score26

    AI hardware sector rebounds across the board: SK Hynix earnings spark rally, Zhongji Innolight up nearly 6.5%

    On October 28, the AI hardware sector rebounded significantly, with CPO, memory chips, PCB, AI chips and other concepts all rising. The ChiNext Index surged 3%, Zhongji Innolight gained nearly 6.5%, and its half-day trading volume exceeded 12 billion yuan. The catalyst came from SK Hynix's third-quarter revenue of 104.62 trillion won and operating profit of 83.33 trillion won, a year-on-year surge of 631%; the company expects the memory supply shortage to persist until 2030. Several brokerages believe the strong outlook for AI computing power remains unchanged, the ChiNext Index valuation sits at the 17.8th percentile of the past 10 years, and opportunities from unjustified sell-offs are emerging.

    #Industry news#English text is being prepared.

  4. East Money
    AI score25

    A-share three major indices rebound, ChiNext Index up 3%, robot industry chain erupts

    The three major A-share indices rebounded collectively, with the ChiNext Index up 3% at 2624.64 points. A total of 4,498 stocks across the market rose, with 80 hitting the limit up, and half-day turnover in the two markets reached 1,013.6 billion yuan. Robot industry chain sectors such as reducers, humanoid robots, and robot actuators led the gains. A Guotai Haitong research report said Tesla's Optimus has started trial production of prototypes, with the first batch of robots meeting mass-production standards officially rolling off the line, and Musk expects external deliveries to begin by the end of 2027.

    #English text is being prepared.#Embodied AI

  5. English text is being prepared.
    AI score58

    Trump says he will formally ban the term AI in relevant documents, renaming it SI Superintelligence

    Trump posted on Truth Social on the 27th, saying that he will soon formally prohibit the use of the term artificial intelligence (AI) in all relevant documents and other occasions, on the grounds that the expression is inaccurate, but did not explain how the ban would be implemented. Previously, on September 29, he signed an executive order requiring all executive departments and agencies to replace AI with superintelligence (SI); Reuters reported on October 7 that the U.S. Department of Justice had issued instructions to internal staff to use superintelligence in the vast majority of occasions, including court documents.

    Same story · 5 reports

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

  6. IT Home · Artificial IntelligenceOpinion
    AI score51

    Anthropic hiring manager says coding interviews are dead, team scraps traditional coding questions

    Felix Rieseberg, Anthropic's hiring manager and head of engineering for Claude-related products, said he has removed traditional coding-question interviews from his team's hiring process, on the grounds that Claude can now easily handle small React challenges and even low-level assembly problems, so coding questions measure AI's abilities more than a candidate's judgment, product taste, and collaboration skills. He also advised junior engineers not to wait for complete requirements documents and cross-department approvals, but instead to target a specific business goal and spend a weekend building a working single-point prototype to win organizational support, summing up this logic as "hunger beats seniority."

    Same story · 5 reports

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

  7. IT Home · Artificial Intelligence
    AI score38

    Research: Always relying on AI to do the work may be quietly degrading your professional skills

    Theoretical philosophy researcher Kaisa Kerki published "Who Should Be Responsible for Addressing Professional Skill Degradation?" in the journal Human Affairs, pointing out that after doctors, researchers, and developers hand their work over to AI, their own professional skills decline, and practitioners find it hard to notice this degradation. She argues that this is a social and structural problem and cannot be entirely blamed on individuals, and recommends that senior practitioners regularly undergo assessments of core competencies, with industry communities defining the core skills that must be mastered by humans.

    Same story · 2 reports

    #Papers/Research#Safety / alignment

  8. The Beijing News
    AI score38

    iFlytek Releases Spark X2.6: How AI Maintains the Boundary of Assistance in Education, Healthcare, and Legal Scenarios

    At the 2026 Global 1024 Developer Festival, iFlytek released the device-cloud integrated iFlytek Spark X2.6 and demonstrated its implementation in livelihood scenarios such as education, healthcare, and legal services. iFlytek Starlight has saved 300,000 registered teachers 60% to 80% of their administrative time, and the Spark Intelligent Grading Machine M50 can grade a class in as fast as 5 minutes; iFlytek Xiaofa, based on the Spark legal large model, showed in trials with nearly 200 lawyers and legal affairs personnel that case-handling efficiency improved by more than 45%. On the healthcare side, iFlytek Xiaoyi has completed over 220 million AI consultations in total, and its positioning has always been a health assistant rather than a diagnostic decision-maker.

    Same story · 5 reports

    #Industry news#Agents#Speech

  9. Wallstreetcn
    AI score78

    Anthropic acquires one-person pharma company Mithril Medicine and partners with six institutions to advance AI drug discovery

    Anthropic's life sciences team has established research collaborations with six institutions—Generate Biomedicines, Divergence Labs, BigHat Biosciences, Enveda, The Column Group, and Bristol Myers Squibb—and has acquired Mithril Medicine, a biotech company independently run by Cody Tranbarger, to directly conduct drug R&D. Anthropic says Claude has delivered quantifiable improvements in molecular design, manufacturing document review, regulatory filings, and clinical trial planning; for example, Enveda used Claude to review 1,061 pages of documents, which identified all 11 issues found by humans and flagged 7 additional ones, cutting manual review time from over 25 hours to under 3 hours. Anthropic says it plans to significantly scale up this work, and if the candidate drugs perform well, it expects to be ready to begin human clinical trials as early as the end of 2028.

    Same story · 5 reports

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

  10. Wall Street Insights
    AI score35

    Morgan Stanley Quant Team: AI Chip Stock Decline Not Over, Recommends Trimming Positions on Rallies

    Morgan Stanley's Ukyo Haraguchi quantitative strategy team believes the selloff in semiconductor and data center-related stocks is not over, with downside risks remaining in the medium term, and recommends investors use rallies to increase their position reductions. The team compared the current trajectory of the Philadelphia Semiconductor Index to the 2000 dot-com bubble period and recommended a "barbell strategy": on one end, allocating to infrastructure bottleneck companies such as power suppliers and select semiconductor names; on the other, allocating to early-stage software enablers and AI application companies.

    #Industry news#English text is being prepared.

  11. Phoenix Technology
    AI score80

    OpenAI Announces Three Results in Mathematics, Computer Science, and Physics; Academia Questions Release Method

    OpenAI announced three breakthroughs in the fields of mathematics, computer science, and physics, and said it will adjust the way it releases the results. The mathematics result is said to prove the Kakeya conjecture in all dimensions; the computer science result claims to disprove the exponential time hypothesis; the physics result is not yet public and has been left to an external expert team to decide the release format. Researchers interviewed by The Verge acknowledged the weight of the results but were dissatisfied that OpenAI selected the experts and restricted access; the Advisory Group on Mathematics and Artificial Intelligence (AGMAI) called for the immediate release of the natural-language manuscript and Lean formal proof of the physics result.

    Same story · 17 reports

    #Industry news#OpenAI#Papers/Research

  12. QbitAIModel releases
    AI score60

    SoMark Team Open-Sources 1.3B Document Parsing Model OXR, Topping Three Benchmarks

    The SoMark team has open-sourced the document parsing model OXR. The main model has only 1.3B parameters and topped all three benchmark tests: OmniDocBench, olmOCR-Bench, and DocParsingBench. Its full parsing speed exceeds that of PaddleOCR-VL-1.6 and MinerU2.5-Pro, which were also tested in the same round. The model covers 13 types of document elements, including text, tables, formulas, chemical structure formulas, and code blocks. It can parse reading order and output layout coordinates, converting PDFs and document images into Markdown / JSON for downstream processes such as Agents and RAG.

    Same story · 2 reports

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

  13. QbitAI
    AI score21

    Damao Technology Proposes E-AIDC Energy-Driven Intelligent Computing Center New Paradigm at CPE 2026

    At CPE 2026, Damao Technology proposed the E-AIDC (energy-driven intelligent computing center) new paradigm, advocating that an energy brain define computing-power-electricity synergy, breaking away from the net electricity value game, and seeking optimization through the integration of four markets: electricity energy, flexibility, capacity, and carbon. Simulation of a project in a western province shows that its energy large model completed calculations for 12,000 scenarios, uncovering a comprehensive value space of more than 3 cents, with investor IRR increasing by about 20% and user delivered electricity prices falling by 10%.

    Same story · 2 reports

    #Industry news#English text is being prepared.

  14. QbitAIProduct release
    AI score60

    Hi3D open-sources the commercially usable 3D model V1.0 and launches a 3D creation Agent

    Hi3D has open-sourced the V1.0 high-resolution geometry refinement model, now available on GitHub and Hugging Face and commercially usable; the model is based on the Sparc3D architecture, with training data of about 2 million filtered 3D assets, 4 times the 500,000 setting in the Sparc3D paper. According to its technical report, V1.0 achieves multi-view fidelity and detail richness of 0.75 and 0.841 respectively, with a slicing success rate of 90%. The simultaneously launched Hi3D Agent is built on Hi3D's professional 3D models as its foundation, connecting intent understanding, generation, modification, and delivery; the latest V3.0 model is claimed to offer commercially usable 2048³ precision.

    Same story · 2 reports

    #Model releases#Open-source ecosystem#Multimodal

  15. QbitAIModel releases
    AI score86

    OpenAI releases GPT-6.1 Terra: API price halved, focused on computer operation capabilities

    OpenAI has released GPT-6.1 Terra, with standard API pricing at $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens—about half the price of GPT-5.6 Terra. OpenAI says that on Agents' Last Exam, it scores 15 percentage points higher than GPT-5.6 Terra at the medium setting, with per-task cost about one-third of the latter, and matches GPT-6 Astra's performance at one-tenth of its token price; on Terminal-Bench 4.0, it solves 50.3% of tasks at the max setting, compared with 21.5% for the previous generation. Plus, Pro, Enterprise, and Edu users can try it in ChatGPT Work and Codex, and developers can call it via gpt-6.1-terra. On the same day, OpenAI also announced three results in mathematics, computer science, and physics produced by its internal AI system, among which the proof concerning the Exponential Time Hypothesis (ETH) has been formally verified in Lean.

    #Model releases#OpenAI#Agents

  16. QbitAIModel releases
    AI score73

    Baidu Open-Sources ERNIE Unlimited OCR 2, Peak KV Cache Reduced from 47GB to 1.77GB

    Baidu has released and open-sourced ERNIE Unlimited OCR 2. Through the sparse reference sliding window attention mechanism SR-SWA and an infinite decoding mechanism, it reduced the peak KV cache from 47GB under standard decoding to 1.77GB in a single-sequence test at the hundred-page level, a reduction of about 96%. In a 16-way concurrency test, end-to-end throughput increased from 360 to 1315 tokens/second, about 3.65 times that of standard decoding. The model has been open-sourced on Hugging Face and GitHub; the previous generation, Unlimited OCR, has accumulated more than 10 million downloads and about 28,600 GitHub stars.

    Same story · 5 reports

    #Model releases#Open-source ecosystem#Multimodal

  17. QbitAIProduct release
    AI score65

    Zhiyuan Robotics releases embodied Agent framework Genie Master, achieving Physical RSI in a convenience store

    Zhiyuan Robotics has released its embodied Agent framework Genie Master, enabling the Genie G2 robot to autonomously analyze the causes of failures, adjust strategies, and accumulate experience while performing tasks in a real convenience store, without relying on manual demonstrations or debugging. The framework connects the upper-level Agent with the underlying action policies through the Embodied Execution Protocol (EEP), and follows two routes—Harness RSI and Policy RSI—to accumulate experience and update action policies respectively. Zhiyuan claims this is the industry's first case of achieving Physical RSI in a real business scenario, breaking away from simulation environments and laboratory settings. On November 2, it will launch a 24-hour convenience store livestream, with 6 Genie G2 robots working consecutive shifts.

    Same story · 5 reports

    #Product updates#Embodied AI#Agents

  18. QbitAIProduct release
    AI score58

    Qianli Technology, under Yin Qi, releases a 4D millimeter-wave radar with a built-in 10 TOPS NPU

    Qianli Technology has released a 4D millimeter-wave radar with a built-in 10 TOPS NPU, handing the entire chain from antenna design to point cloud output over to the model, and supporting OTA upgrades of perception capabilities. Compared with the previous generation of 4D radar, which used DSP-based on-device computing, point cloud density is enhanced 10-fold, detection range is increased by 40%, ghost elimination is improved to 95%, and spatial angular resolution is improved 12-fold. The radar offers three operating modes: edge-side, satellite architecture, and compatible; under the satellite architecture, the raw data input volume may be about 100 times that of traditional 4D radar, with ADC bandwidth reaching 6 Gbps.

    #Product updates#English text is being prepared.#English text is being prepared.

  19. QbitAIProduct release
    AI score71

    RoboParty Releases Fully Open-Source Humanoid Robot RP1

    RoboParty has released the fully open-source humanoid robot RP1, with the scope of openness covering mechanical drawings, electrical systems, the self-developed ROMOMO joints, low-level control, simulation training, and the deployment pipeline. RP1 stands 135cm tall and offers two body configurations: an open-source version with 24 active degrees of freedom and an open-source Pro version with 30 degrees of freedom. It has demonstrated real-machine capabilities including continuous walking, human motion tracking, disturbance recovery, and getting up after a fall. Resources such as the URDF, MJCF, and mesh models have already been made public, along with the open-sourced hhtools human motion retargeting tool; some materials, including CAD, BOM, and circuit board manufacturing files, remain internal engineering records, and the specific scope of openness is subject to the project repository.

    #Product updates#Open-source ecosystem#Embodied AI

  20. Shanghai ObserverOpinion
    AI score57

    Shanghai Observer commentary: After AI face-swapping breaks through facial verification, the governance bottleneck lies on the institutional side

    A Shanghai Observer commentary points out that an AI face-swapping personal information protection public interest litigation case heard by the Hangzhou Intermediate People's Court this year shows that AI face-swapping can already break through facial verification; the Ministry of Public Security reported that in the first nine months of 2026, more than 300 AI face-swapping fraud cases were solved nationwide, with the amount involved exceeding 500 million yuan. The article holds that current sector-specific regulations each govern one segment but cannot govern the whole chain, that platform labeling obligations are hard-pressed to stop fraud, that the governance bottleneck lies in incentives and penalties on the institutional side, and recommends that comprehensive legislation make the very provision of AI generation tools that bypass security verification an independently punishable act.

    #English text is being prepared.#Safety / alignment#English text is being prepared.