Skip to content

All AI news

Today 8 items
Oct 7Wed
  1. The Decoder59

    South Korea bets $3.49 billion on building a homegrown frontier AI model to rival China's best

    South Korea plans to invest 4.7 trillion won (about $3.49 billion) through government equity investment to develop a homegrown frontier AI model; the funding is part of the 2027 budget proposal and still requires National Assembly approval. The new plan expands roughly ninefold on the previous 530 billion won given to five companies and sets up a second track to support Korean models' deployment; the government will launch an open competition that startups can also join, rather than automatically increasing the bids of LG AI Research, SK Telecom and Upstage, and officials say Korea aims to match China's leading open-source models.

  2. OpenAI News72

    Building advertising for the way people use AI

    OpenAI announced new visual advertising formats in ChatGPT, along with expanded measurement tools for advertisers, attribution partnerships and brand suitability capabilities.

    Why it matters: OpenAI has officially announced visual ad formats in ChatGPT plus expanded ad measurement, attribution and brand suitability tools, making its monetization direction worth watching.

  11. OpenAI News66

    Sharing AI progress in mathematics

    OpenAI released new results on its internal frontier model solving open mathematical problems, and published Lean proof formalizations and research details on GitHub.

    Why it matters: The article reports that OpenAI shared results from its internal frontier model on open math problems and released Lean formalized proofs and research details on GitHub.

  16. Microsoft Research48

    What AI gets wrong and what failure teaches us

    In a Microsoft Research podcast, partner research manager Jennifer Neville speaks with applied scientist Chad Atalla about how evaluation drives AI systems past performance limits to meet user needs, and about the "unexpected failures" models show when tested beyond standard benchmarks. She shares practical advice on using existing AI systems and stresses the importance of carefully checking data when results don't match expectations.

  17. MIT Technology Review · AI54

    People really hate AI, so why can’t they get enough?

    MIT Technology Review's Will Douglas Heaven analyzes the public's love-hate relationship with AI: a Pew survey shows US adults think AI's negatives outweigh its positives, and a May Gallup poll found 71% of US adults oppose building new AI data centers locally. Yet ChatGPT reached 1 billion monthly users by May, Gemini hit 950 million users in July, and Pew says half of US adults use chatbots.

  18. Google Research47

    Open and Emergent Problems in Agentic Privacy and Security: A Contextual Angle

    Google released a report from its CAPS workshop, bringing together more than 50 academic and industry leaders to discuss the privacy and security challenges of autonomous agents. The report notes that agents differ from traditional software in three ways—unstructured interfaces, probabilistic control flow, and autonomous delegation—and, building on Contextual Integrity theory, proposes contextual policy engines combined with system-level sandboxing, model-level reasoning, and user-centric controls as a multi-layered defense.

  19. MIT Technology Review · AI21

    Bringing predictive analytics to the agentic AI era

    The frontier of enterprise AI is shifting from prediction to autonomous decision-making, and the core question for 2026 is how to let predictive systems act on their conclusions without drifting from business intent. Deep learning and generative AI enable real-time training so models keep evolving, and predictive engines' data now extends to unstructured interaction sources. Everest Group partner Vishal Gupta says enterprises are no longer satisfied with retrospective views, and "the word analytics is giving way to AI—everything is becoming AI."

  25. TechCrunch · AI55

    TikTok rolls out an AI shopping assistant and one-click checkout

    TikTok announced on Monday a conversational AI shopping assistant and in-app one-click checkout, letting users order from brands directly in the For You feed. The assistant understands context, remembers user preferences, and offers real-time shopping guidance on product details, shipping, sizing, and inventory while helping complete purchases; the new features were built in partnership with e-commerce and payments platforms including Salesforce, Shopify, Shoplazza, and Stripe.

  26. TechCrunch · AI45

    Instinct brings its AI agent to group chats, even for friends without an account

    AI agent company Instinct (valued at $10 billion after its latest funding round) launched a group chat feature that lets users add Instinct to group chats with friends for trip planning, ticket sniping, carpooling, and more, with friends able to use it without registering an Instinct account. The feature is rolling out to early access users first and will expand to all users 'soon'. On privacy, group chat Instinct is isolated from users' personal accounts, and personal agents need user permission before sharing information or taking actions.

  30. Simon Willison58

    Quoting Felix Rieseberg

    Anthropic's Felix Rieseberg described the new Cowork architecture: the old version ran tool calls in a local VM with model inference in the cloud, but users disliked the disk, battery and performance overhead and it stopped when the laptop lid closed. The new version moves both model inference and the VM to the cloud, with each session in an isolated sandbox sharing no state; when user device files (such as local files) need access, the desktop app handles that file access tool call, enabling phone use, tasks that keep running, and lower battery drain.

  32. AWS Machine Learning Blog56

    Agentic retrieval with LangChain and Amazon Bedrock Knowledge Bases

    AWS demonstrated on Amazon Bedrock Managed Knowledge Bases with LangChain a comparison of standard versus agentic retrieval: for a multi-part question containing six intents, single vector retrieval at 5 results covered only 4 sub-intents, while agentic retrieval used a planning loop to break the query into sub-queries, evaluate evidence and iterate.

  33. Simon Willison41

    Scrimshaw Jukebox

    Simon Willison tested whether Claude Opus 5.5 can compose music: he had it design a simple text-based music format, build a playable artifact, and generate sample tracks benchmarked against the soundtrack quality of Secret of Monkey Island. The results were surprisingly good, though the style leaned heavily Monkey Island. The author speculates that musical composition may be a newly emergent capability of text models in recent months, and says experiments with other new and older models are needed to verify it.

  37. Simon Willison18

    datasette-atom 0.11a0

    Simon Willison released datasette-atom 0.11a0 on October 6. It is a pre-release of the Datasette plugin, and npm-related statistics show the datasette tag now has 1,549 associated projects.

  40. Simon Willison23

    Mistral Large 4

    Simon Willison tested Mistral Large 4 with an SVG generation prompt about an armadillo in fishnet stockings jaywalking on Mars, comparing its output with claude-opus-5.5, gpt-6.1-sol, and gemini-3.8-flash, with each model using its default reasoning level.