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
2mowish I knew! from the post it seems like an LLM without a decoder, RL'd on multiple choice questions, but more details would be coolwish I knew! from the post it seems like an LLM without a decoder, RL'd on multiple choice questions, but more details would be cool
↗Not briefed
rss
simon-willison
2moGemini Live audio—
↗Not briefed
bluesky
bluesky-ai
2mo「うわこの書き方LLMだろ、どうせハルシネってるやろなぁ」「うわこの書き方LLMだろ、どうせハルシネってるやろなぁ」
↗Not briefed
hn
hn-ai
2moAnthropic Confirms Claude AI Use in Weapons and Surveillance—
↗Not briefed
bluesky
bluesky-ai
2mo割と困っているというかアレなんですけど、LLMに書かせたであろう文章に拒否感が出てきて最近そういった文章臭いと分かった瞬間何も読めなくなる現象が頻発している割と困っているというかアレなんですけど、LLMに書かせたであろう文章に拒否感が出てきて最近そういった文章臭いと分かった瞬間何も読めなくなる現象が頻発している
↗Not briefed
arxiv
arxiv-ai
2moWalking the Score Manifold: Continuous-time Generative Dynamics on Learned Data ManifoldsGenerative modeling of time-dependent data is typically formulated on a discrete temporal grid, restricting supervision to the observe…
↗Not briefed
bluesky
bluesky-ai
2moOnce again, what is intrinsic to LLM technology besides the massive energy waste is that they will degrade in an infinite loop of regressio…Once again, what is intrinsic to LLM technology besides the massive energy waste is that they will degrade in an infinite loop of regr…
↗Not briefed
bluesky
bluesky-ai
2mo"we can't possibly model weather accurately, it's too computationally expensive! It'd leave out too many variables!""we can't possibly model weather accurately, it's too computationally expensive! It'd leave out too many variables!"
↗Not briefed
arxiv
arxiv-ai
2moQEMScore: How Much Does the Measurement Add to Learned Quantum Error Mitigation?How much does the noisy measurement add to learned quantum error mitigation? An accuracy table cannot say, because a model handed circ…
↗Not briefed
bluesky
bluesky-ai
2momaybe this is why they teach three-cueing so your brain is just making probabilistic guesses at what things mean like an LLM doing the samemaybe this is why they teach three-cueing so your brain is just making probabilistic guesses at what things mean like an LLM doing th…
↗Not briefed
arxiv
arxiv-ai
2moTabPFN-3.5: Technical ReportWe introduce TabPFN-3.5, our new flagship Tabular Foundation Model. It significantly outperforms its predecessor, TabPFN-3, and all ex…
↗Not briefed
bluesky
bluesky-ai
2moTotally! My current hypothesis is he wants some fame/truly believes it (for now). Interestingly, the only *real info* he’s given the public…Totally! My current hypothesis is he wants some fame/truly believes it (for now). Interestingly, the only *real info* he’s given the p…
↗Not briefed
bluesky
bluesky-ai
2mooh, also, worth noting: Thiel is heavily involved in the whole arc. he was an early funder of MIRI and Yudkowsky, and is still invested in …oh, also, worth noting: Thiel is heavily involved in the whole arc. he was an early funder of MIRI and Yudkowsky, and is still investe…
↗Not briefed
arxiv
arxiv-ai
2moBracketing Uncertainty in Clustering Under the Manifold HypothesisThe manifold hypothesis suggests a natural criterion for clustering: partition data according to the manifold component from which eac…
↗Not briefed
bluesky
bluesky-ai
2moThere is a direct, traceable line between Yudkowsky's work and the publications of MIRI, through to LessWrong and the rationalist movement,…There is a direct, traceable line between Yudkowsky's work and the publications of MIRI, through to LessWrong and the rationalist move…
↗Not briefed
bluesky
bluesky-ai
2moYou can shoot me for realizing this but “it’s not just this – it’s THAT” LLM syntax is the basic compositional principle behind the text of…You can shoot me for realizing this but “it’s not just this – it’s THAT” LLM syntax is the basic compositional principle behind the te…
↗Not briefed
arxiv
arxiv-ai
2moOBC-Prune: Outcome-Based Calibration for Large Reasoning Model PruningLarge reasoning models (LRMs) generate long chain-of-thought traces before answering, creating significant inference overhead. Pruning…
↗Not briefed
bluesky
bluesky-ai
2moA push by top AI firms to coordinate with one another and the US government on safeguards threatens to make it harder for small companies t…A push by top AI firms to coordinate with one another and the US government on safeguards threatens to make it harder for small compan…
↗Not briefed
rss
techmeme
2moTypeSafe AI, which is developing a model that outputs numerical responses with probability estimates to help businesses assess reliability, raised a $40M seed (Rashi Shrivastava/Forbes)Rashi Shrivastava / Forbes: TypeSafe AI, which is developing a model that outputs numerical responses with probability estimates to he…
↗Not briefed
arxiv
arxiv-ai
2moLong-Context Demonstration Selection Using State Space Models🔧 ReplicateWe study the problem of demonstration selection, which involves selecting a subset of examples for prepending to a query to a language…
↗Not briefed
bluesky
bluesky-ai
2moThis is extremely intriguing and might portend a future where LLMs perform many of the tasks they're currently good at vastly more efficien…This is extremely intriguing and might portend a future where LLMs perform many of the tasks they're currently good at vastly more eff…
↗Not briefed
bluesky
bluesky-ai
2moOne of my visual aids is this map which I hope illustrates how groupthink can do easily form but I'm not sure I'm getting all the way throu…One of my visual aids is this map which I hope illustrates how groupthink can do easily form but I'm not sure I'm getting all the way…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI backs bills in US Congress on AI and biological weapon threats reut.rs/4y7GN3hOpenAI backs bills in US Congress on AI and biological weapon threats reut.rs/4y7GN3h
↗Not briefed
arxiv
arxiv-ai
2moDataset-Dependent Effects of Cross-Depth Aggregation and Soft-Routed Experts in EEG Foundation Model Fine-TuningEEG decoding tasks can rely on different temporal dynamics and cross-channel relationships. We test whether specialized modules improv…
↗Not briefed
hn
hn-ai
2moMove AI chats across harnesses: Claude Code, Codex, OpenCode, Cursor, and more—
↗Not briefed
bluesky
bluesky-ai
2moA related but slightly different issue is the way that "AI safety" can and will be used to justify neo-colonialism. We've already seen ma…A related but slightly different issue is the way that "AI safety" can and will be used to justify neo-colonialism. We've already se…
↗Not briefed
hn
hn-ai
2moAdios MCP – let coding agents build, preview, debug and deploy apps—
↗Not briefed
arxiv
arxiv-ai
2moERPBench: A State-Grounded Evaluation Paradigm for Computer-Use Agents in Enterprise SoftwareComputer-use agents that operate through screenshots and simulated actions are advancing rapidly, yet their evaluation remains anchore…
↗Not briefed
arxiv
arxiv-ai
2moThe Unbearable Weight: Scaling Models and Methods for UAV Audio ClassificationAs unmanned aerial vehicles (UAVs) become increasingly prevalent in consumer and defense settings, classifying them reliably from limi…
↗Not briefed
hn
hn-ai
2moAustralian government paper proposes opt-out for AI training—
↗Not briefed
arxiv
arxiv-ai
2moDoes AI Assistance Leave a Temporal Fingerprint? Detecting Overreliance in AI-Assisted Writing and ProgrammingThe rapid adoption of generative AI has made final artifacts unreliable evidence of student learning, and AI detectors that examine on…
↗Not briefed
bluesky
bluesky-ai
2mo現時点でAIが「できる」ことを脅威とみなして批判する人もいるし、「できない」ことを看板倒れとして批判する人もいるわけですが: AIを使わないことをポリシーとされている方々の「AIは○○ができない」というご主張の一部は、AI開発競争の進展とともに実態とズレ始めている印象がありま…現時点でAIが「できる」ことを脅威とみなして批判する人もいるし、「できない」ことを看板倒れとして批判する人もいるわけですが: AIを使わないことをポリシーとされている方々の「AIは○○ができない」というご主張の一部は、AI開発競争の進展とともに実態とズレ始めている印…
↗Not briefed
bluesky
bluesky-ai
2moWould he still have OpenAI money from his previous employer? He’s also a very young guy who pretty likely believes this stuff and would hav…Would he still have OpenAI money from his previous employer? He’s also a very young guy who pretty likely believes this stuff and woul…
↗Not briefed
bluesky
bluesky-ai
2moGLM kinda goated at fleshing out procgen… of the cheap models at least. Wonder how Gemini 3.8 flash doesGLM kinda goated at fleshing out procgen… of the cheap models at least. Wonder how Gemini 3.8 flash does
↗Not briefed
arxiv
arxiv-ai
2moCan VLMs Reliably Assess Sidewalk Accessibility Attributes from Pedestrian-Level Imagery?An important component of urban accessibility, particularly for wheelchair users and people with reduced mobility, is sidewalk complia…
↗Not briefed
hn
hn-ai
2moLAUSD Adopts Moratorium on Student AI Use, Another Move to Limit Classroom Tech—
↗Not briefed
rss
techmeme
2moSources: OpenAI has held early conversations with investors about raising another private funding round at a $1.2T valuation before its planned IPO (Financial Times)Financial Times: Sources: OpenAI has held early conversations with investors about raising another private funding round at a $1.2T va…
↗Not briefed
hn
hn-ai
2moWhat I measured after Hacker News broke my LLM agent gate—
↗Not briefed
bluesky
bluesky-ai
2moOpenAI weighs funding round at $1.2tn valuation before IPO ft.trib.al/zAN6fqfOpenAI weighs funding round at $1.2tn valuation before IPO ft.trib.al/zAN6fqf
↗Not briefed
bluesky
bluesky-ai
2moThe problem is that getting from here to step one requires, essentially, magic. And if you point that out to one of them, they lose their s…The problem is that getting from here to step one requires, essentially, magic. And if you point that out to one of them, they lose th…
↗Not briefed
bluesky
bluesky-ai
2moThey don't seem to reveal how it's implemented but Gemini guesses they have taken some open source LLM, stripped away the autoregressive to…They don't seem to reveal how it's implemented but Gemini guesses they have taken some open source LLM, stripped away the autoregressi…
↗Not briefed
bluesky
bluesky-ai
2moThe people who are closest to this shit and those who use and interact with it the most all inevitably seem to develop psychosis to one deg…The people who are closest to this shit and those who use and interact with it the most all inevitably seem to develop psychosis to on…
↗Not briefed
hn
hn-ai
2moLean Verified Transformers—
↗Not briefed
hn
hn-ai
2moAsk HN: Slow OpenAI Inference on AWS BedrockInference on AWS Bedrock for OpenAI models has been very slow and glitchy the last few days with today being the worst. Very long late…
↗Not briefed
rss
techmeme
2moAt a US House hearing, Treasury Secretary Scott Bessent said AI labs should get no liability exemptions and called for more open-source models built in the US (Matt Bracken/FedScoop)Matt Bracken / FedScoop: At a US House hearing, Treasury Secretary Scott Bessent said AI labs should get no liability exemptions and c…
↗Not briefed
hn
hn-ai
2moGitLost: We Tricked GitHub's AI Agent into Leaking Private Repos—
↗Not briefed
bluesky
bluesky-ai
2moPart of the logic is that they think so little of *other* people's humanity that they think the other people could be replicated with an LL…Part of the logic is that they think so little of *other* people's humanity that they think the other people could be replicated with…
↗Not briefed
bluesky
bluesky-ai
2moLLM couldn't give me SVG code for an inverted triangle. It couldn't invert the triangle it gave me. It can't do concepts or mathematical tr…LLM couldn't give me SVG code for an inverted triangle. It couldn't invert the triangle it gave me. It can't do concepts or mathematic…
↗Not briefed
bluesky
bluesky-ai
2moMon hypothèse c'est que c'est du marketing "regardez comment mon LLM il est tellement balèze qu'il pourrait bien tous nous tuer... Lol..." …Mon hypothèse c'est que c'est du marketing "regardez comment mon LLM il est tellement balèze qu'il pourrait bien tous nous tuer... Lol…
↗Not briefed
bluesky
bluesky-ai
2moTbh, I kind of doubt that we can count on the "AI" industry being destroyed by people running it being convinced it will kill everyone, bec…Tbh, I kind of doubt that we can count on the "AI" industry being destroyed by people running it being convinced it will kill everyone…
↗Not briefed