←Search results

By relevance · Total 2778 results Clear search

Thursday

  1. China.com TechnologySelected

    Google Releases Gemini 4 Carbon Large Model, with a Focus on Enhancing Cybersecurity Capabilities

    Google has launched its new-generation large model Gemini 4 Carbon, focusing on cybersecurity and long-horizon complex task handling, and it ranks first with an overall score of 59 on the Artificial Analysis Intelligence Index v5.0. On WebDev Arena, the model improved by 89 Elo points over the previous generation to 1769 points, ranking fourth; internal testing shows that its success rate in outputting secure code is more than twice that of Gemini 4 Argon. Pricing is initially the same as Gemini 4 Argon: during the promotional period, input is $2 per million tokens and output is $10 per million tokens. This week it is being made available for trial to a small number of cybersecurity practitioners on a priority basis.

    Google has launched its new-generation large model Gemini 4 Carbon, focusing on cybersecurity and long-horizon complex task handling, and it ranks first with an overall score of 59 on the Artificial Analysis Intelligence Index v5.0. On WebDev Arena, the model improved by 89 Elo points over the previous generation to 1769 points, ranking fourth; internal testing shows that its success rate in outputting secure code is more than twice that of Gemini 4 Argon. Pricing is initially the same as Gemini

    Matched fieldsBody

  2. Hugging Face

    AnySpeechRecognizer and AnyDecisionModel: Open-source speech and decision models for the Apple platform

    AnySpeechRecognizer supports engine switching on top of the Apple SpeechAnalyzer API, with options including NVIDIA Parakeet TDT 0.6B v3 (supporting 25 European languages), Moondream post-trained Parakeet Ultra, the English version Parakeet Unified, and Alibaba Qwen3-ASR (0.6B/1.7B, covering 30 languages and 22 Chinese dialects). These models have been converted to Core ML and published on the Hugging Face Hub; Parakeet takes about 20–40 seconds to transcribe 1 hour of audio, while Qwen3-ASR is slower and requires several GB of memory. The package also adds a speaker diarization module, SpeakerDiarizer, which the Apple framework lacks; it is based on NVIDIA Streaming Sortformer and can track up to 4 speakers.

    AnySpeechRecognizer supports engine switching on top of the Apple SpeechAnalyzer API, with options including NVIDIA Parakeet TDT 0.6B v3 (supporting 25 European languages), Moondream post-trained Parakeet Ultra, the English version Parakeet Unified, and Alibaba Qwen3-ASR (0.6B/1.7B, covering 30 languages and 22 Chinese dialects). These models have been converted to Core ML and published on the Hugging Face Hub; Parakeet takes about 20–40 seconds to transcribe 1 hour of audio, while Qwen3-ASR is

    Matched fieldsBody

  3. Hugging Face

    Hugging Face Announces Project ProtoStar: Making Publicly Funded Scientific Data AI-Ready

    Hugging Face announced Project ProtoStar to promote the systematic sharing of publicly funded scientific research outputs in AI-ready formats, accompanied by open infrastructure and release guidelines. Hugging Face already hosts more than 17,000 public datasets in core scientific fields, including nearly 29,000 in medicine and mathematics; the number of new scientific datasets added in the first nine months of 2026 is more than triple that of the same period in 2025. The first challenge focuses on climate risk, using historical weather data from 210 U.S. cities through 2018 to predict actual climate conditions from 2019 to 2026.

    Hugging Face announced Project ProtoStar to promote the systematic sharing of publicly funded scientific research outputs in AI-ready formats, accompanied by open infrastructure and release guidelines. Hugging Face already hosts more than 17,000 public datasets in core scientific fields, including nearly 29,000 in medicine and mathematics; the number of new scientific datasets added in the first nine months of 2026 is more than triple that of the same period in 2025. The first challenge focuses

    Matched fieldsBody

  4. Meta:Newsroom · AI

    Meta Partners with Five U.S. Universities to Accelerate Medical and Scientific Breakthroughs Using Its Models

    Meta is partnering with five U.S. universities—Caltech, Ohio State University, Louisiana State University, Indiana University, and UTHealth Houston—providing $2 million in funding and access to its latest models to support research in medicine, physics, chemistry, biology, computer science, and mathematics using its models. Specific projects include using the open-weight SAM model to monitor tracheostomy tube displacement in infants, using multimodal AI to help dementia patients communicate, and using open-weight models combined with abdominal CT, medical records, and ECG data to identify HFpEF heart failure risk early.

    Meta is partnering with five U.S. universities—Caltech, Ohio State University, Louisiana State University, Indiana University, and UTHealth Houston—providing $2 million in funding and access to its latest models to support research in medicine, physics, chemistry, biology, computer science, and mathematics using its models. Specific projects include using the open-weight SAM model to monitor tracheostomy tube displacement in infants, using multimodal AI to help dementia patients communicate, and

    Matched fieldsBody

  5. AIhub

    From Teaching Machines to Teaching People: An Interview with Michael Littman

    In AIhub's AI Pioneers interview series, reinforcement learning researcher Michael Littman looks back on his journey from building his own reinforcement learning system with tic-tac-toe in the 1980s to only learning in 1988 that this counted as reinforcement learning. He believes that deep reinforcement learning has solved many behavioral problems by throwing computing power at them, but its data demands are too great and it struggles to make creative leaps the way animals do, while real-world scenarios such as robot navigation remain far from solved. He also discusses the challenges generative AI poses to education, and science communication that spreads computer science concepts through music.

    In AIhub's AI Pioneers interview series, reinforcement learning researcher Michael Littman looks back on his journey from building his own reinforcement learning system with tic-tac-toe in the 1980s to only learning in 1988 that this counted as reinforcement learning. He believes that deep reinforcement learning has solved many behavioral problems by throwing computing power at them, but its data demands are too great and it struggles to make creative leaps the way animals do, while real-world s

    Matched fieldsBody

  6. Sina Tech

    Driven by AI financing boom, global bond issuance hits record high

    So far this year, global investment-grade and junk bond issuance has raised a total of $6.61 trillion, up 15% year-on-year, surpassing the 2025 record with two months still left in the year. Among this, cloud computing giants and data center developers issued nearly $368 billion in bonds, more than six times the same period last year, accounting for about one-third of the increase in borrowing. Amazon became the world's fourth-largest bond issuer and largest corporate borrower this year, with Alphabet close behind; five of the top ten U.S. corporate issuers are tech companies.

    So far this year, global investment-grade and junk bond issuance has raised a total of $6.61 trillion, up 15% year-on-year, surpassing the 2025 record with two months still left in the year. Among this, cloud computing giants and data center developers issued nearly $368 billion in bonds, more than six times the same period last year, accounting for about one-third of the increase in borrowing. Amazon became the world's fourth-largest bond issuer and largest corporate borrower this year, with Al

    Matched fieldsBody

  7. Ed Zitron long-form article / Where's Your Ed At

    Ed Zitron's Long-Form Article Questions Anthropic IPO: S-1 Conceals Q3 Data, $430 Billion Compute Commitment About to Take Effect

    In his long-form article, Ed Zitron analyzes Anthropic's S-1, arguing that the document only discloses Q2 2026 data, and that the quarter's $974 million in profit came from two months of SpaceX compute discounts, while Q3 revenue, cash balance, and payment schedule were not disclosed. He tallies Anthropic's non-cancelable compute commitments with Google, Amazon, Microsoft, SpaceX, Broadcom, and others at approximately $269.4 billion, with an additional $125 billion to $150 billion in TPU purchase commitments not included in the contract table, and notes that Q3 revenue grew 45% quarter-over-quarter to $16.8 billion, while compute spending grew 76.3% to $9.88 billion. The article also mentions that 5% to 7% of revenue comes from customers found to have violated rules by using its models to train their own models, as well as lock-up arrangements that allow insiders to sell down relatively quickly.

    In his long-form article, Ed Zitron analyzes Anthropic's S-1, arguing that the document only discloses Q2 2026 data, and that the quarter's $974 million in profit came from two months of SpaceX compute discounts, while Q3 revenue, cash balance, and payment schedule were not disclosed. He tallies Anthropic's non-cancelable compute commitments with Google, Amazon, Microsoft, SpaceX, Broadcom, and others at approximately $269.4 billion, with an additional $125 billion to $150 billion in TPU purchas

    Matched fieldsBody

  8. Cartesia Product and Research

    Cartesia Launches Managed Agents End-to-End Voice Agent Platform

    Cartesia has launched the Managed Agents end-to-end voice agent platform, targeting enterprise call scenarios. On Artificial Analysis's Speech Agent Arena, it ranks first among 30 agents with a preference Elo of 1115. On ServiceNow's EVA-Bench, the platform achieved an accuracy of 87.4% and an experience score of 90.7%, both the highest. The platform co-designs the voice model with the harness, offering capabilities such as 44 languages, tool calling, knowledge bases, and outbound calling batches. The company says it has already supported tens of millions of conversations in finance, retail, hospitality, and healthcare scenarios.

    Cartesia has launched the Managed Agents end-to-end voice agent platform, targeting enterprise call scenarios. On Artificial Analysis's Speech Agent Arena, it ranks first among 30 agents with a preference Elo of 1115. On ServiceNow's EVA-Bench, the platform achieved an accuracy of 87.4% and an experience score of 90.7%, both the highest. The platform co-designs the voice model with the harness, offering capabilities such as 44 languages, tool calling, knowledge bases, and outbound calling batche

    Matched fieldsBody

Wednesday

  1. AIhub

    Why AI Is Difficult to Move from the Lab to the Real World: Three Major Challenges of Generalization, Human Behavior, and Resources

    When AI vision systems move from the lab to homes, hospitals, and streets, they face three major challenges: generalization, human behavior, and resources. Most models are trained only on well-lit, high-quality images, and their human pose estimation performance drops noticeably in low-light or nighttime scenes. Using an unsupervised domain adaptation method, the authors' team, in a 2026 study on low-light human pose estimation, substantially improved model performance on two low-light benchmarks without requiring manual joint annotations for low-light images. The number of action combinations in human-object interaction is enormous and difficult to collect and annotate one by one, so the research instead uses large vision-language models to recognize new interactions unseen during training.

    When AI vision systems move from the lab to homes, hospitals, and streets, they face three major challenges: generalization, human behavior, and resources. Most models are trained only on well-lit, high-quality images, and their human pose estimation performance drops noticeably in low-light or nighttime scenes. Using an unsupervised domain adaptation method, the authors' team, in a 2026 study on low-light human pose estimation, substantially improved model performance on two low-light benchmark

    Matched fieldsBody

  2. Tessl: Product and Engineering Blog

    Who bears the cost of the AI software factory?

    Tessl proposes managing the cost of the AI software factory through seven stages: cost visibility and attribution, task admission assessment, selecting models based on the total cost of completing a task, replacing repeated model decisions with deterministic checks, budget boundaries, value accounting, and accountability. The article argues that every model call, tool operation, and workflow should attribute its cost to the task that triggered it, and failed retries should also be counted; otherwise, counting only successful tasks would make the process appear cheaper than it actually is.

    DevCon NYC

    Tessl proposes managing the cost of the AI software factory through seven stages: cost visibility and attribution, task admission assessment, selecting models based on the total cost of completing a task, replacing repeated model decisions with deterministic checks, budget boundaries, value accounting, and accountability. The article argues that every model call, tool operation, and workflow should attribute its cost to the task that triggered it, and failed retries should also be counted; other

    Matched fieldsBody

  3. 404 Media

    Meta updates its commercial terms to allow using user content to train AI; UK and EU users can object

    Meta updated its commercial terms this month to allow it to use content from associated domains, web pages, or URLs to develop, improve, research, train, and maintain Meta products, sparking discontent among artists and small business owners who rely on Instagram and Facebook. The effective date was postponed from October 30 to November 20, and Meta subsequently removed the word "train" from the list of permitted uses and limited the uses to "advertising and business products." UK and EU users can currently submit an objection form in the Instagram Privacy Center, and some users say they have received emails from Meta confirming that their objections were approved; these options do not appear in the United States and other places.

    Meta updated its commercial terms this month to allow it to use content from associated domains, web pages, or URLs to develop, improve, research, train, and maintain Meta products, sparking discontent among artists and small business owners who rely on Instagram and Facebook. The effective date was postponed from October 30 to November 20, and Meta subsequently removed the word "train" from the list of permitted uses and limited the uses to "advertising and business products." UK and EU users c

    Matched fieldsBody

  4. 404 Media

    UK and EU users can object to Meta using their content to train AI

    Meta updated its terms for business accounts to allow it to use content from linked websites and pages to develop, improve, and train AI, sparking dissatisfaction among small businesses and artists who use Instagram. UK and EU users can currently submit an objection form in the Instagram Privacy Center, and some users say they have received emails from Meta approving their objections. The terms were originally set to take effect on October 30 but were changed to November 20 after objections, and Meta also removed the word "train" from the permitted uses and narrowed the scope to advertising and commercial products.

    Meta updated its terms for business accounts to allow it to use content from linked websites and pages to develop, improve, and train AI, sparking dissatisfaction among small businesses and artists who use Instagram. UK and EU users can currently submit an objection form in the Instagram Privacy Center, and some users say they have received emails from Meta approving their objections. The terms were originally set to take effect on October 30 but were changed to November 20 after objections, and

    Matched fieldsBody

Friday

  1. Baseten Engineering BlogSelected

    Baseten releases decision model joey-1, with P50 latency of 20ms and $0.01 per million tokens

    Baseten's Base Labs has released joey-1, a 2B decision model that directly outputs a single-token label and calibrated confidence, with a P50 end-to-end latency of 20ms, about 2.5 times faster than elephant-1 (49ms), at a price of $0.01 per million tokens, one-tenth that of mammoth-1. joey-1 is post-trained from Qwen3.5-2B and shares the same data recipe and API with elephant-1 and mammoth-1, so switching requires changing only one line. The model is now available on Baseten Model APIs, with zero data retention by default, and its weights are open-sourced on Hugging Face; Model APIs supports a 16k context window, while dedicated deployments can unlock a 262k context window and support fine-tuned weights.

    joey-1

    Baseten's Base Labs has released joey-1, a 2B decision model that directly outputs a single-token label and calibrated confidence, with a P50 end-to-end latency of 20ms, about 2.5 times faster than elephant-1 (49ms), at a price of $0.01 per million tokens, one-tenth that of mammoth-1. joey-1 is post-trained from Qwen3.5-2B and shares the same data recipe and API with elephant-1 and mammoth-1, so switching requires changing only one line. The model is now available on Baseten Model APIs, with zer

    Matched fieldsBody

  2. Sina Tech

    34.5% of SpaceX's Q3 Revenue Came from Anthropic Compute Contract

    According to people familiar with the matter, Anthropic is SpaceX's single largest customer, contributing more than one-third of its revenue in the third quarter. SpaceX documents show that this customer accounts for about 34.5% of revenue, and the documents label it only as "Customer B." This revenue comes from Anthropic's use of compute at SpaceX's data center in Memphis, Tennessee. The contract is worth up to $120.5 billion and runs through 2029, but starting this December Anthropic will pay about $3.1 billion per month and may terminate the contract with 90 days' notice after January 2027 and March 2027.

    According to people familiar with the matter, Anthropic is SpaceX's single largest customer, contributing more than one-third of its revenue in the third quarter. SpaceX documents show that this customer accounts for about 34.5% of revenue, and the documents label it only as "Customer B." This revenue comes from Anthropic's use of compute at SpaceX's data center in Memphis, Tennessee. The contract is worth up to $120.5 billion and runs through 2029, but starting this December Anthropic will pay

    Matched fieldsBody

  3. Sina Tech

    Musk predicts again: superintelligence will bring super abundance, making it meaningless to save for the future

    Musk posted on X that with superintelligence ushering in an era of extreme material abundance, it is meaningless to save money for the distant future. He previously suggested on a podcast that there is no need to save for retirement 10 or 20 years from now, and predicted that advances in AI, energy, and robotics will bring abundant resources, allowing everyone to have everything they want. He also proposed that the government issue subsidies to all citizens to cope with job losses caused by AI.

    Matched fieldsBody

  4. Wallstreetcn

    Amazon Q3 Earnings: AWS Revenue Up 39.3%, AI Capital Expenditures Turn Free Cash Flow Negative

    Amazon's net sales in Q3 2026 grew 12% year-over-year to $201.94 billion, slightly below expectations, while AWS net sales grew 39.3% year-over-year to $45.98 billion, the highest single-quarter growth rate in nearly five years. Purchases of property and equipment for the quarter reached $77.26 billion, about 1.2 times the year-ago period, and free cash flow swung from an inflow of $14.79 billion in the year-ago period to an outflow of $42.36 billion. The company maintained its full-year 2026 cash capital expenditure guidance of approximately $220 billion unchanged, and the stock fell as much as more than 10% in after-hours trading following the earnings release.

    Amazon's net sales in Q3 2026 grew 12% year-over-year to $201.94 billion, slightly below expectations, while AWS net sales grew 39.3% year-over-year to $45.98 billion, the highest single-quarter growth rate in nearly five years. Purchases of property and equipment for the quarter reached $77.26 billion, about 1.2 times the year-ago period, and free cash flow swung from an inflow of $14.79 billion in the year-ago period to an outflow of $42.36 billion. The company maintained its full-year 2026 ca

    Matched fieldsBody

  5. Wallstreetcn

    Anthropic Reportedly SpaceX's Largest Single Customer, Contributing Over One-Third of Q3 Revenue

    According to Bloomberg, citing people familiar with the matter, Anthropic is currently SpaceX's largest single customer, contributing more than one-third of SpaceX's revenue in the third quarter, exceeding the share from the U.S. federal government. SpaceX documents show that about 34.5% of revenue that quarter came from a single AI division customer, "Customer B." Anthropic has committed to paying SpaceX up to $120.5 billion over the contract term through 2029, equivalent to about $3.1 billion per month starting this December, but it can terminate the contract with 90 days' notice after January and March 2027.

    According to Bloomberg, citing people familiar with the matter, Anthropic is currently SpaceX's largest single customer, contributing more than one-third of SpaceX's revenue in the third quarter, exceeding the share from the U.S. federal government. SpaceX documents show that about 34.5% of revenue that quarter came from a single AI division customer, "Customer B." Anthropic has committed to paying SpaceX up to $120.5 billion over the contract term through 2029, equivalent to about $3.1 billion

    Matched fieldsBody

  6. Wall Street Insights

    Microsoft's $25 Billion Bond Offering Draws 4.6x Subscription, Yet 30-Year Spread Exceeds Lower-Rated P&G

    Microsoft completed a $25 billion bond offering, with peak orders exceeding $115 billion and a subscription ratio of 4.6 times, across 10 tranches maturing in 2 to 40 years. However, its 30-year bond spread reached 95 basis points above the benchmark Treasury, higher than the 30-year bond issued the same week by lower-rated P&G (60 basis points). Hyperscalers and data center developers have raised more than $313 billion in global investment-grade bond financing this year, nearly three times higher than the full year of 2025.

    Microsoft completed a $25 billion bond offering, with peak orders exceeding $115 billion and a subscription ratio of 4.6 times, across 10 tranches maturing in 2 to 40 years. However, its 30-year bond spread reached 95 basis points above the benchmark Treasury, higher than the 30-year bond issued the same week by lower-rated P&G (60 basis points). Hyperscalers and data center developers have raised more than $313 billion in global investment-grade bond financing this year, nearly three times high

    Matched fieldsBody

  7. Wallstreetcn

    Nvidia Plans to Provide Credit Guarantee for Nscale's Over $5 Billion Data Center Construction Financing

    Nvidia is in talks with emerging AI cloud service provider Nscale to provide over $5 billion in credit guarantees for the debt financing of a 460-megawatt data center Nscale plans to build in West Virginia, USA; the agreement has not yet been signed. The data center will serve Anthropic, with the two parties having signed a $45 billion lease contract, and the facility will be equipped with Nvidia Vera Rubin GPUs. The guarantee takes the form of a residual value guarantee and covers the data center facility itself rather than the lease or the chips. To date, Nvidia has provided approximately $170 billion in credit support and investment to customers including emerging cloud providers.

    Nvidia is in talks with emerging AI cloud service provider Nscale to provide over $5 billion in credit guarantees for the debt financing of a 460-megawatt data center Nscale plans to build in West Virginia, USA; the agreement has not yet been signed. The data center will serve Anthropic, with the two parties having signed a $45 billion lease contract, and the facility will be equipped with Nvidia Vera Rubin GPUs. The guarantee takes the form of a residual value guarantee and covers the data cent

    Matched fieldsBody

  8. Wallstreetcn

    Amazon Earnings Call: AWS Expected to Grow Over 40% in 2027 and 2028, 2026 Capex Around $220 Billion

    Amazon CEO Andy Jassy said on the Q3 earnings call that AWS third-quarter revenue grew 39.3% year over year to $45.98 billion, and is expected to grow over 40% in both 2027 and 2028, becoming a business with annual revenue exceeding one trillion dollars in the next 5 to 10 years. The company maintained its expectation of about $220 billion in cash capital expenditure for 2026, most of which will go to AWS, and expects spending to increase significantly in 2027. On Bedrock, spending by Fortune 500 companies grew nearly 7x year over year and total tokens grew nearly 17x, revenue commitments for Trainium and Graviton grew more than 20x year over year, and Trainium is planned to be sold externally by the end of 2027.

    Amazon CEO Andy Jassy said on the Q3 earnings call that AWS third-quarter revenue grew 39.3% year over year to $45.98 billion, and is expected to grow over 40% in both 2027 and 2028, becoming a business with annual revenue exceeding one trillion dollars in the next 5 to 10 years. The company maintained its expectation of about $220 billion in cash capital expenditure for 2026, most of which will go to AWS, and expects spending to increase significantly in 2027. On Bedrock, spending by Fortune 50

    Matched fieldsBody