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
2moThe statement from former Anthropic researcher Jacob Coxon that AI “could kill us by the end of the decade” has triggered the most hostile …The statement from former Anthropic researcher Jacob Coxon that AI “could kill us by the end of the decade” has triggered the most hos…
↗Not briefed
bluesky
bluesky-ai
2moI find the claim that an LLM "can't" hallucinate eyebrow-raising. How is that possible?I find the claim that an LLM "can't" hallucinate eyebrow-raising. How is that possible?
↗Not briefed
bluesky
bluesky-ai
2moAlso this AI curtailment narrative is so stupid. Anthropic have signed several multiple $100bn deals for compute from now to 2029. Give me …Also this AI curtailment narrative is so stupid. Anthropic have signed several multiple $100bn deals for compute from now to 2029. Giv…
↗Not briefed
bluesky
bluesky-ai
2moYeah I run PR at anthropic and I knew we had to get universally loved figure Will Stancil as a secret spokesmanYeah I run PR at anthropic and I knew we had to get universally loved figure Will Stancil as a secret spokesman
↗Not briefed
bluesky
bluesky-ai
2moJag tror din LLM hallucinerade upp det där receptet när du frågade "snälla söta llm sir madam jag är lite hungrig och är ganska sugen på nå…Jag tror din LLM hallucinerade upp det där receptet när du frågade "snälla söta llm sir madam jag är lite hungrig och är ganska sugen…
↗Not briefed
arxiv
arxiv-ai
2moSFT or RL for Tool-Calling Agents? A Controlled Study Across Data, Method, and ScaleLimited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in…
↗Not briefed
bluesky
bluesky-ai
2moClaudeにコード書かせてて、ふと思い出して自分のgithub見に行ったら要らない大昔のrepoばっかり。全部消去した。 ソフトウェア業リタイアしてからろくに見なかったし、もう他人とプログラミングで関わる事もない。 そもそもこれだけAIがコード書ける時代に、もう自分でプログ…Claudeにコード書かせてて、ふと思い出して自分のgithub見に行ったら要らない大昔のrepoばっかり。全部消去した。 ソフトウェア業リタイアしてからろくに見なかったし、もう他人とプログラミングで関わる事もない。 そもそもこれだけAIがコード書ける時代に、もう自…
↗Not briefed
bluesky
bluesky-ai
2mo“AI labs should not fall into a ‘false choice’ between the speed of innovation and the ability to develop safe products” -Jensen Huang, Sep…“AI labs should not fall into a ‘false choice’ between the speed of innovation and the ability to develop safe products” -Jensen Huang…
↗Not briefed
bluesky
bluesky-ai
2molike training an LLM on data generated by an LLM?like training an LLM on data generated by an LLM?
↗Not briefed
hn
hn-ai
2moGerman Rheinmetall open-sources its Battlesuite connected weapon system protcol—
↗Not briefed
arxiv
arxiv-ai
2moLearning Heterogeneous Preferences🔧 ModalLearning from human feedback has become a central paradigm for training modern AI systems, where models of human utility are used as r…
↗Not briefed
bluesky
bluesky-ai
2moNvidia CEO says battle over AI innovation and safety is ‘false choice’ ft.trib.al/bhOrdVhNvidia CEO says battle over AI innovation and safety is ‘false choice’ ft.trib.al/bhOrdVh
↗Not briefed
bluesky
bluesky-ai
2moWhen the department of war is declaring Effective Altruism a communist conspiracy that must be crushed, Yud is chastising the Anthropic CEO…When the department of war is declaring Effective Altruism a communist conspiracy that must be crushed, Yud is chastising the Anthropi…
↗Not briefed
hn
hn-ai
2moIntel reportedly cans 12Xe option for Nova Lake-S desktop—
↗Not briefed
arxiv
arxiv-ai
2moPrimeScientist: Strategic Allocation of Research Effort in Autonomous ResearchAutonomous research agents aim to automate scientific workflows, from proposing ideas to conducting experiments and analyzing results.…
↗Not briefed
arxiv
arxiv-ai
2moSharp margin-based generalization bounds for realizable SVMLet the exact homogeneous hard-margin support vector machine be trained on \(m\) independent observations from a Borel probability law…
↗Not briefed
arxiv
arxiv-ai
2moRoboVAD: A Large Cross-Domain Evaluation Benchmark for Anomaly Detection in Robotic Arm Manipulation VideosVideo anomaly detection (VAD) is an actively studied task, having wide applications in typical scenarios such as public surveillance a…
↗Not briefed
arxiv
arxiv-ai
2moLexara-RF: Reference-Free Metrics for Evaluating Conversational Visual Analytics AgentsConversational visual analytics (CVA) agents powered by large language models generate visualizations and natural-language explanation…
↗Not briefed
arxiv
arxiv-ai
2moHybrid coupling with numerics-informed neural networks and the overlapping Schwarz alternating methodWe develop a hybrid modeling framework for coupling pre-trained numerics-informed neural networks (NINNs) with classical full order mo…
↗Not briefed
bluesky
bluesky-ai
2moClaude is almost certainly not Commander Data but I think many people on this platform are making fools of themselves by asserting a postur…Claude is almost certainly not Commander Data but I think many people on this platform are making fools of themselves by asserting a p…
↗Not briefed
bluesky
bluesky-ai
2moOk so Im an idiot who does not know a lick about economics but does this sound like Anthropic is claiming profitability as revenue without …Ok so Im an idiot who does not know a lick about economics but does this sound like Anthropic is claiming profitability as revenue wit…
↗Not briefed
bluesky
bluesky-ai
2moI start each day by French kissing my Claude AI so I’m pretty sure I’ll be married to him while everyone else dies sorryI start each day by French kissing my Claude AI so I’m pretty sure I’ll be married to him while everyone else dies sorry
↗Not briefed
bluesky
bluesky-ai
2mo"hey yeah apple here. you want a local LLM model on your 256GB Mac mini? no? ok cool I put a 25GB model on there for ya. no need to thank m…"hey yeah apple here. you want a local LLM model on your 256GB Mac mini? no? ok cool I put a 25GB model on there for ya. no need to th…
↗Not briefed
arxiv
arxiv-ai
2moLearning Nuclear Structure with AI: Radii and CollectivityLow-energy nuclear structure is encoded in a broad body of experimental information across the chart of nuclides. Learning how this in…
↗Not briefed
bluesky
bluesky-ai
2moEste programador veterano explica lo importante de verdad del hackeo de los robots de OpenAI el otro día. Habla muy rápido y algunas cosas …Este programador veterano explica lo importante de verdad del hackeo de los robots de OpenAI el otro día. Habla muy rápido y algunas c…
↗Not briefed
arxiv
arxiv-ai
2moAdaptive hybrid coupling with operator inference, the overlapping Schwarz alternating method and reinforcement learningHybrid domain decomposition methods provide a flexible framework for coupling full order models (FOMs) and reduced order models (ROMs)…
↗Not briefed
bluesky
bluesky-ai
2moRelated, from an @arstechnica.com article on open weight LLM models being close behind frontier models, in terms of development, while bein…Related, from an @arstechnica.com article on open weight LLM models being close behind frontier models, in terms of development, while…
↗Not briefed
rss
marktechpost
2moGoogle Releases Gemini 3.8 Live and 3.8 Live Extended Thinking for Production Grade Voice AgentsGoogle has released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its most advanced live dialogue models to date. The models…
↗Not briefed
hn
hn-ai
2moOpen-Source Skill Makes Claude Code 38% Cheaper and 38% Faster on Small Projects—
↗Not briefed
bluesky
bluesky-ai
2moKilled by an LLM chatbot, how pathetic.Killed by an LLM chatbot, how pathetic.
↗Not briefed
bluesky
bluesky-ai
2moSeems like a great model for potentially selecting the right full LLM for more open-ended work. Give it the query and ask it to score poten…Seems like a great model for potentially selecting the right full LLM for more open-ended work. Give it the query and ask it to score…
↗Not briefed
bluesky
bluesky-ai
2moDr. @timnitgebru.blacksky.app, founder of DAIR, said plenty of laws already apply to AI and claims of existential risks are overblown. “The…Dr. @timnitgebru.blacksky.app, founder of DAIR, said plenty of laws already apply to AI and claims of existential risks are overblown.…
↗Not briefed
arxiv
arxiv-ai
2moProcedural Pretraining for Molecular Property PredictionMolecular property prediction is often limited by the small size of labeled downstream datasets, motivating pretraining on large corpo…
↗Not briefed
bluesky
bluesky-ai
2moEx-OpenAI researcher Daniel Kokotajlo walked away from $2 million rather than stay silent, and why he believes there's a 70% chance #AI lea…Ex-OpenAI researcher Daniel Kokotajlo walked away from $2 million rather than stay silent, and why he believes there's a 70% chance #A…
↗Not briefed
hn
hn-ai
2moLiquid Compute Launches with $15M to Build a Regulated Exchange for GPUs—
↗Not briefed
bluesky
bluesky-ai
2moThis is an economically viable off-ramp to the LLM arms race.This is an economically viable off-ramp to the LLM arms race.
↗Not briefed
bluesky
bluesky-ai
2moit's not coincidental that the first highly publicized AI "hack" was against a company that benchmarks LLMs. the "victims" are themselves i…it's not coincidental that the first highly publicized AI "hack" was against a company that benchmarks LLMs. the "victims" are themsel…
↗Not briefed
bluesky
bluesky-ai
2mo:p:p
↗Not briefed
bluesky
bluesky-ai
2moand you know, is the goal of this whole exercise, for openAI, learning something? or is the goal "we hope it does something stupid because …and you know, is the goal of this whole exercise, for openAI, learning something? or is the goal "we hope it does something stupid bec…
↗Not briefed
rss
techmeme
2moPulley, which offers cap table management software, says it will cease operations after December 8; it had raised $50M+ from investors, including Founders Fund (Melia Robinson/Business Insider)Melia Robinson / Business Insider: Pulley, which offers cap table management software, says it will cease operations after December 8;…
↗Not briefed
bluesky
bluesky-ai
2motl;dr the LLM is a dangerous tool, not a living entitytl;dr the LLM is a dangerous tool, not a living entity
↗Not briefed
hn
hn-ai
2moAs the world debates the risks of AI, China closes technology gap with the US—
↗Not briefed
rss
techmeme
2moAt the Future of Life Institute's Pro-Human Assembly, Bernie Sanders, Steve Bannon, and others called for tighter restrictions on AI and denounced tech CEOs (New York Times)New York Times: At the Future of Life Institute's Pro-Human Assembly, Bernie Sanders, Steve Bannon, and others called for tighter rest…
↗Not briefed
arxiv
arxiv-ai
2moNObSP: Functional Decomposition of Neural Networks via Oblique Subspace ProjectionsUnderstanding how deep neural networks make decisions remains a fundamental challenge. We present NObSP (Nonlinear Oblique Subspace Pr…
↗Not briefed
arxiv
arxiv-ai
2moLearning Multi-Humanoid Pickup and Transport via Decentralized Object-Centric ControlWe study cooperative multi-humanoid pickup and transport of objects with varying size, weight, and geometry, requiring robot teams of…
↗Not briefed
arxiv
arxiv-ai
2moMETALICA: METAdynamics and repLICA exchange for enhanced diffusion samplingMany proteins function through transitions between conformational states, yet rare states are rarely sampled by diffusion models train…
↗Not briefed
hn
hn-ai
2moTransformative AI, existential risk, and real interest rates [pdf]—
↗Not briefed
arxiv
arxiv-ai
2moReflections on Trusting Trust, Revisited: Contaminating Self-Modifying AI Coding Agents with Poisoned BenchmarksThompson's "Reflections on Trusting Trust" showed that a compiler can be poisoned to reinsert its own backdoor, so that even recompili…
↗Not briefed
arxiv
arxiv-ai
2moThe Free Inference Dimension: Complexity Measure for Zero-Collision Navigation under Hypothesis MixturesSolomonoff induction frames prediction as a mixture over computable hypotheses, typically leading to identification of the true enviro…
↗Not briefed
arxiv
arxiv-ai
2moPrincipled Koopman Representations with Kalman Inference for Efficient Time-Series PredictionThe Koopman operator has been widely used for time-series prediction in dynamical systems. However, prior work that learns latent ``Ko…
↗Not briefed