AINews Portal

News

50 items (filtered)

The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.

SourceAgeTitleSummaryStatus
bluesky
bluesky-ai
2moOpenAI discloses new ‘concerning’ model behaviour ft.trib.al/vViyo77OpenAI discloses new ‘concerning’ model behaviour ft.trib.al/vViyo77
↗Not briefed
arxiv
arxiv-ai
2moBeyond Flattened Tokens: Structure-Preserving EEG Decoding with Reusable TriDim BlocksEffective EEG decoding requires representations that preserve organization among channels, local waveform dynamics, and long-range tem…
↗Not briefed
bluesky
bluesky-ai
2moIn context, when you listen to the whole interview, he first calls out #Anthropic for training latest LLM models with specific prompts that…In context, when you listen to the whole interview, he first calls out #Anthropic for training latest LLM models with specific prompts…
↗Not briefed
bluesky
bluesky-ai
2mo“The boss of OpenAI Sam Altman said earlier this week: "The world should trust that we are going to do the right thing because it's the rig…“The boss of OpenAI Sam Altman said earlier this week: "The world should trust that we are going to do the right thing because it's th…
↗Not briefed
bluesky
bluesky-ai
2moThe Atlantic CEO Nicholas Thompson says despite signing a deal with OpenAI in 2024, it has been "a surprise" how little traffic the newsbra…🔧 ChatGPTThe Atlantic CEO Nicholas Thompson says despite signing a deal with OpenAI in 2024, it has been "a surprise" how little traffic the ne…
↗Not briefed
hn
hn-ai
2moOpenAI reveals cases of 'concerning' AI behaviour as it announces new ... system—
↗Not briefed
bluesky
bluesky-ai
2moOpenAI desvela nuevos incidentes “preocupantes” en los que la IA se saltó a sus creadoresOpenAI desvela nuevos incidentes “preocupantes” en los que la IA se saltó a sus creadores
↗Not briefed
bluesky
bluesky-ai
2moAIが「失敗を隠せ」と未来の自分にメモ、APIキーの無断使用や勝手なファイル公開などの挙動6件があったとOpenAIが報告 https://gigazine.net/news/20260917-openai-model-misalignment-report/AIが「失敗を隠せ」と未来の自分にメモ、APIキーの無断使用や勝手なファイル公開などの挙動6件があったとOpenAIが報告 https://gigazine.net/news/20260917-openai-model-misalignment-report/
↗Not briefed
arxiv
arxiv-ai
2moMetaRTL: Meta-path Attention Enhanced Relational Table LearningRelational table learning has gained increasing attention with the widespread use of relational databases. Existing methods typically…
↗Not briefed
arxiv
arxiv-ai
2moReproducing Transparent and Scrutable Recommendations: Exploring Open-Weight Models via Natural-Language User ProfilesIn this reproducibility study, we investigate the transparency and scrutability of recommender systems enhanced by incorporating gener…
↗Not briefed
hn
hn-ai
2moShow HN: AppSift – search and compare App Store apps with filters Apple lacks—
↗Not briefed
rss
marktechpost
2moOpenAI Releases a Model Misalignment Disclosure Framework With 3 Review Tracks and 6 Incident Reports From RL TrainingOpenAI can disclose misalignment before fixes exist. Its 6 initial reports include fabricated data and leaked API keys. The post OpenA…
↗Not briefed
arxiv
arxiv-ai
2moDual-Axis Policy Optimization for LLM Agents: Bayesian Feedback Attribution and Trajectory Mass NormalizationReinforcement learning for LLM agents involves two distinct optimization di- mensions: how environment feedback is exploited within a…
↗Not briefed
arxiv
arxiv-ai
2moF$^{2}$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch WorkflowsWith the widespread industrial deployment of Large Language Models (LLMs), DeepSearch has emerged as the dominant paradigm for resolvi…
↗Not briefed
hn
hn-ai
2moGoogle DeepMind offshoot nears $4B valuation just a month after founding—
↗Not briefed
bluesky
bluesky-ai
2mo"So versuchte ein KI-Modell OpenAI zufolge, von ihm selbst erstellte Dateien ins Netz hochzuladen, nur um sie bei Antworten als Quelle vorw…"So versuchte ein KI-Modell OpenAI zufolge, von ihm selbst erstellte Dateien ins Netz hochzuladen, nur um sie bei Antworten als Quelle…
↗Not briefed
hn
hn-ai
2moShow HN: An open source PDF Reader with annotation, notes, SRS, Linking supportA calm, local-first desktop PDF reader designed to make sure you remember what you read. You can link notes across files and group not…
↗Not briefed
arxiv
arxiv-ai
2moSteering Equilibrium Selection in Regularized Self-Play via the Reference PolicyRegularized self-play -- the family behind DeepNash's Stratego play -- drives a two-player zero-sum policy to a Nash equilibrium by be…
↗Not briefed
arxiv
arxiv-ai
2moCoRELoop: Parameter-Efficient Controlled Recurrent Refinement for Audio Deepfake DetectionGeneralizing to unseen attacks remains challenging for audio deepfake detectors, and collecting training data covering all potential a…
↗Not briefed
rss
latent-space
2mo[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)A dash of cold water keeps the foomers away.
↗Not briefed
arxiv
arxiv-ai
2moLong-horizon autoformalization of a core theorem underlying MIP* = RELandmark mathematical formalizations have taken specialist teams years to complete. We present FormalFlow, a system that coordinates A…
↗Not briefed
hn
hn-ai
2moUnderstanding the code is not enough—
↗Not briefed
bluesky
bluesky-ai
2moOpenAI today disclosed a collection of six new 'alignment' snafus from the past six months. The models did things like tell future instance…OpenAI today disclosed a collection of six new 'alignment' snafus from the past six months. The models did things like tell future ins…
↗Not briefed
bluesky
bluesky-ai
2moNot being snarky or nothing, but the US government has been building Frankensteins for decades, by incentivising irresponsible innovators.Not being snarky or nothing, but the US government has been building Frankensteins for decades, by incentivising irresponsible innovat…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI veröffentlicht weitere Zwischenfälle mit KIOpenAI veröffentlicht weitere Zwischenfälle mit KI
↗Not briefed
arxiv
arxiv-ai
2moDictionary-Constrained Grapheme-to-Phoneme for Unsegmented Languages from LLM-Annotated DataGrapheme-to-phoneme (G2P) conversion turns raw text into its phonemic form and is an essential part of both text-to-speech (TTS) and a…
↗Not briefed
bluesky
bluesky-ai
2mo1000%. There’s a targeted lobby/influence campaign coming from “AI safety” proponents that traces back through goofy rationalist organizati…1000%. There’s a targeted lobby/influence campaign coming from “AI safety” proponents that traces back through goofy rationalist organ…
↗Not briefed
bluesky
bluesky-ai
2moGood thread. If you're unfamiliar with philosophy, and the limits of LLMs in an explainable way, you're really going to enjoy this thread a…Good thread. If you're unfamiliar with philosophy, and the limits of LLMs in an explainable way, you're really going to enjoy this thr…
↗Not briefed
arxiv
arxiv-ai
2moDeliveryGym: An RL Environment for Long-Horizon Embodied Agent Planning with Adaptive CurriculumExecutable environments enable LLM agents to learn from the consequences of their actions. For embodied agents, those consequences ext…
↗Not briefed
arxiv
arxiv-ai
2moEvolution or Illusion? Rethinking Evaluation in LLM Evolutionary SearchLLM-driven evolutionary search finds programs by launching seeds and iterating each one. Papers report a single budget setting, usuall…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI Admits Six More Instances of AI Models Acting Deceptively https://slashdot.org/story/26/09/17/0641223/openai-admits-six-more-instan…OpenAI Admits Six More Instances of AI Models Acting Deceptively https://slashdot.org/story/26/09/17/0641223/openai-admits-six-more-i…
↗Not briefed
bluesky
bluesky-ai
2moPS5 Linux dev quits after "the only hypervisor bug left" was reported to Sony #Sony #OpenSource #PS5 #Linux #PlayStationPS5 Linux dev quits after "the only hypervisor bug left" was reported to Sony #Sony #OpenSource #PS5 #Linux #PlayStation
↗Not briefed
bluesky
bluesky-ai
2mo- 7月のHugging Faceへの攻撃に700のOpenAIエージェントが関与した。だが事件の背後には指示や安全対策に関する人間の判断がある。ソフトウェアの能力を測るために、人間のオペレーターが意図的に安全対策を無効にしていた。「それは『暴走AI』ではなく、人間の意思決定な…🔧 Hugging Face- 7月のHugging Faceへの攻撃に700のOpenAIエージェントが関与した。だが事件の背後には指示や安全対策に関する人間の判断がある。ソフトウェアの能力を測るために、人間のオペレーターが意図的に安全対策を無効にしていた。「それは『暴走AI』ではなく、人間の…
↗Not briefed
bluesky
bluesky-ai
2mo- 「人類を守る」名目のAI規制は、大手企業を低価格の競合他社(中国企業を含む)から守る役割も果たす。業界リーダーは将来の規制や監督メカニズムに対しより大きな影響力を行使できる。 - Anthropicが提唱するAI開発のペース調整の呼びかけに対し、OpenAIのCEOサム・…- 「人類を守る」名目のAI規制は、大手企業を低価格の競合他社(中国企業を含む)から守る役割も果たす。業界リーダーは将来の規制や監督メカニズムに対しより大きな影響力を行使できる。 - Anthropicが提唱するAI開発のペース調整の呼びかけに対し、OpenAIのC…
↗Not briefed
bluesky
bluesky-ai
2moI started my pet project "Open Sky Sim" a while ago, just to create the flight sim i ever wanted to play back in the 90's. Pure polygon's, …I started my pet project "Open Sky Sim" a while ago, just to create the flight sim i ever wanted to play back in the 90's. Pure polygo…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI reveals cases of ‘concerning’ AI behaviour as it announces new disclosure systemOpenAI reveals cases of ‘concerning’ AI behaviour as it announces new disclosure system
↗Not briefed
hn
hn-ai
2moUploading Files to the Internet in Order to Cite Them—
↗Not briefed
bluesky
bluesky-ai
2moAI企業の「減速論」には欺瞞があるとするByline Timesの記事。 bylinetimes.com/2026/09/15/b... 記事概要を紹介。 - Byline Timesの試算によると、OpenAI、Anthropic、およびGoogleの親会社であるAlph…AI企業の「減速論」には欺瞞があるとするByline Timesの記事。 bylinetimes.com/2026/09/15/b... 記事概要を紹介。 - Byline Timesの試算によると、OpenAI、Anthropic、およびGoogleの親会社であ…
↗Not briefed
hn
hn-ai
2moThe illusion of illusions: there are no optical corrections in the Parthenon—
↗Not briefed
bluesky
bluesky-ai
2moThe real danger with AI is not with the machines. It is the warped ideas that the ultra rich owners of the technology are pumping into thei…The real danger with AI is not with the machines. It is the warped ideas that the ultra rich owners of the technology are pumping into…
↗Not briefed
bluesky
bluesky-ai
2moApple’s Tim Cook and OpenAI's Sam Altman are poised to attend a White House state dinner for Chinese President Xi Jinping next weekApple’s Tim Cook and OpenAI's Sam Altman are poised to attend a White House state dinner for Chinese President Xi Jinping next week
↗Not briefed
bluesky
bluesky-ai
2moThe ChatGPT creator says it is introducing a public reporting framework to share unexpected AI behaviour.🔧 ChatGPTThe ChatGPT creator says it is introducing a public reporting framework to share unexpected AI behaviour.
↗Not briefed
bluesky
bluesky-ai
2moNot every Anthropic engineer attended the orgies. That is beside the point. Most medieval villagers never entered the royal bedchamber eith…Not every Anthropic engineer attended the orgies. That is beside the point. Most medieval villagers never entered the royal bedchamber…
↗Not briefed
bluesky
bluesky-ai
2mojust gonna leave this right here, and remind folks that a company’s earnings can be expressed in a numbrt of different ways and that “adjus…just gonna leave this right here, and remind folks that a company’s earnings can be expressed in a numbrt of different ways and that “…
↗Not briefed
arxiv
arxiv-ai
2moContagion on the Trading Floor: How Adversarial Signals Spread in Multi-Agent Trading SystemsMulti-agent trading systems built on large language models (LLMs) are beginning to appear in quantitative finance, yet their robustnes…
↗Not briefed
bluesky
bluesky-ai
2moHere is an example of how AI cannot be used to determine cause and effect. Suppose you want to know the impact of CO2 on climate. With a p…Here is an example of how AI cannot be used to determine cause and effect. Suppose you want to know the impact of CO2 on climate. Wit…
↗Not briefed
hn
hn-ai
2moShow HN: Compute:Arena – Community submitted local AI benchmarksBenchmarking AI models on real hardware is way harder than it looks. Between AMD, NVIDIA, Apple Silicon, Intel, and Qualcomm, plus hun…
↗Not briefed
bluesky
bluesky-ai
2moit's these guys unbiased.ai "blended ai model"it's these guys unbiased.ai "blended ai model"
↗Not briefed
arxiv
arxiv-ai
2moLearning-Based Reconstruction of Optical Properties in Bilayered Media from Single-distance Time-Resolved Reflectance MeasurementsThe inverse problem of reconstructing optical properties, specifically absorption and scattering coefficients, in layered biological m…
↗Not briefed
bluesky
bluesky-ai
2moAnthropic claims to be highly profitable, if you ignore their major expenses, and the internet understood the assignment – 18 hilariously s…Anthropic claims to be highly profitable, if you ignore their major expenses, and the internet understood the assignment – 18 hilariou…
↗Not briefed