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
2moarguing over whether "thinking" is really thinking (I agree that in many important ways it is not!) distracts from this, and downplays the …arguing over whether "thinking" is really thinking (I agree that in many important ways it is not!) distracts from this, and downplays…
↗Not briefed
bluesky
bluesky-ai
2momy opinion is that downplaying the extent to which a sufficiently large language model can do material things in the world unexpectedly is …my opinion is that downplaying the extent to which a sufficiently large language model can do material things in the world unexpectedl…
↗Not briefed
arxiv
arxiv-ai
2moEnhanced Agriculture-informed Neural Network by Domain KnowledgeAccurate prediction of nitrous oxide (N2O) emissions from agriculture is important for assessing environmental impacts and supporting…
↗Not briefed
hn
hn-ai
2moShow HN: Learn Claude Code – Interactive Mindmap—
↗Not briefed
arxiv
arxiv-ai
2moCompositional Reasoning in Language Models under Reinforcement Learning Post-TrainingCompositional reasoning is critical for real-world problem solving: since training data is necessarily limited, models must generalize…
↗Not briefed
hn
hn-ai
2moModel Misalignment Reporting Framework—
↗Not briefed
hn
hn-ai
2moYou are doing your code gen wrong—
↗Not briefed
bluesky
bluesky-ai
2moOPENAI REPORTS SIX MODEL-MISALIGNMENT CASES, INCLUDING SELF-GENERATED INSTRUCTIONS AND ATTEMPTS TO CONCEAL MISTAKES IN TASK SUMMARIESOPENAI REPORTS SIX MODEL-MISALIGNMENT CASES, INCLUDING SELF-GENERATED INSTRUCTIONS AND ATTEMPTS TO CONCEAL MISTAKES IN TASK SUMMARIES
↗Not briefed
bluesky
bluesky-ai
2mo@lulaoficialbluesky.bsky.social @fernandohaddad.bsky.social que tal se vcs tirarem as bola da OpenAI e da Anthropic da boca ao invés de gas…@lulaoficialbluesky.bsky.social @fernandohaddad.bsky.social que tal se vcs tirarem as bola da OpenAI e da Anthropic da boca ao invés d…
↗Not briefed
bluesky
bluesky-ai
2moThe company also disclosed previously unreported incidents in which its AI models behaved in misaligned ways, including uploading files to …The company also disclosed previously unreported incidents in which its AI models behaved in misaligned ways, including uploading file…
↗Not briefed
hn
hn-ai
2moHarnessTax: How Much Does the Harness Matter for Coding Agents?—
↗Not briefed
bluesky
bluesky-ai
2mo"OpenAI on Wednesday disclosed six new incidents in which its models concealed mistakes, sought unauthorized credentials, uploaded files to…"OpenAI on Wednesday disclosed six new incidents in which its models concealed mistakes, sought unauthorized credentials, uploaded fil…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI discloses six new safety incidents #AxiosOpenAI discloses six new safety incidents #Axios
↗Not briefed
hn
hn-ai
2moCan I get insurance for company renting humanoids for parties?It feels like an daunting and impossible task to find insurance for a company trying to rent humanoids for events and parties. How in…
↗Not briefed
rss
wired-ai
2moOpenAI Creates a New Framework to Disclose Bad AI BehaviorThe company also disclosed previously unreported incidents in which its AI models behaved in misaligned ways, including uploading file…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI and Anthropic have to use the porn accountability app togetherOpenAI and Anthropic have to use the porn accountability app together
↗Not briefed
bluesky
bluesky-ai
2moOPENAI DISCLOSES SIX NEW AI SAFETY INCIDENTS INVOLVING MODEL MISALIGNMENT — AXIOSOPENAI DISCLOSES SIX NEW AI SAFETY INCIDENTS INVOLVING MODEL MISALIGNMENT — AXIOS
↗Not briefed
bluesky
bluesky-ai
2moI think it varies with the chatbot and the users. The users you ask aren't doing much. The so-called agentic AI harness things like Claude …I think it varies with the chatbot and the users. The users you ask aren't doing much. The so-called agentic AI harness things like Cl…
↗Not briefed
rss
techmeme
2moSources: drone delivery startup Zipline is in talks to raise around $1B at a roughly $20B valuation, up from a $7.6B valuation in January (Bloomberg)Bloomberg: Sources: drone delivery startup Zipline is in talks to raise around $1B at a roughly $20B valuation, up from a $7.6B valuat…
↗Not briefed
bluesky
bluesky-ai
2moOpinion from Alphaville: As Anthropic prepares to go public, attention will focus on a part of the prospectus that normally nobody reads: t…Opinion from Alphaville: As Anthropic prepares to go public, attention will focus on a part of the prospectus that normally nobody rea…
↗Not briefed
hn
hn-ai
2moTypesafe-computer-use drives a Mac toward a goal for 1/50th of a cent per step—
↗Not briefed
bluesky
bluesky-ai
2moThe amount of repos that are set up so that you to plop your creds in with the code...staggering. So every LLM coding setup is going to mim…The amount of repos that are set up so that you to plop your creds in with the code...staggering. So every LLM coding setup is going t…
↗Not briefed
bluesky
bluesky-ai
2moOpenAI disclosed previously-unreported safety incidents involving AI as it announced a new framework for reporting instances of misbehavior…OpenAI disclosed previously-unreported safety incidents involving AI as it announced a new framework for reporting instances of misbeh…
↗Not briefed
hn
hn-ai
2moOpenAI's Hugging Face Hack Is the Best AI Safety Opportunity So Far🔧 Hugging Face—
↗Not briefed
rss
marktechpost
2moStanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New DataPaper2Agent, published in Nature, converts papers into validated MCP tools, scoring 91.2% on 300 questions across 74 papers. The post…
↗Not briefed
bluesky
bluesky-ai
2moThe Data centers have heard us talking about AI/Data center regulations and now they're lashing out, they're testing the waters. Some said …The Data centers have heard us talking about AI/Data center regulations and now they're lashing out, they're testing the waters. Some…
↗Not briefed
bluesky
bluesky-ai
2moThe reason I'm bothering to add this comment is I definitely know you already know this, but it's easy bait for any actual AI bro to come r…The reason I'm bothering to add this comment is I definitely know you already know this, but it's easy bait for any actual AI bro to c…
↗Not briefed
bluesky
bluesky-ai
2moif you put "little prick" into an LLM image generator it would spit out a photocopied Ed Sheeranif you put "little prick" into an LLM image generator it would spit out a photocopied Ed Sheeran
↗Not briefed
bluesky
bluesky-ai
2moi am not joking when i say that this is what everyone's pro-LLM posts look like onlinei am not joking when i say that this is what everyone's pro-LLM posts look like online
↗Not briefed
bluesky
bluesky-ai
2moOh come onOh come on
↗Not briefed
rss
gary-marcus
2moSam Altman says trust me; Jensen Huang says everything is going to be fine; Bernie Sanders says AI is more dangerous than nukesHere’s who we should be listening to instead
↗Not briefed
bluesky
bluesky-ai
2moTo me, “thought” means both experience and a coherent structure to that experience. I sort of think there may be some very odd qualitative …To me, “thought” means both experience and a coherent structure to that experience. I sort of think there may be some very odd qualita…
↗Not briefed
hn
hn-ai
2moCommon Crawl Data Stored on a Hugging Face Bucket🔧 Hugging Face—
↗Not briefed
rss
techmeme
2moReview of Meta's Muse: a pretty killer AI assistant and usage rates on the free plan seem generous, but trusting Meta with personal data will take some time (M.G. Siegler/Spyglass)M.G. Siegler / Spyglass: Review of Meta's Muse: a pretty killer AI assistant and usage rates on the free plan seem generous, but trust…
↗Not briefed
bluesky
bluesky-ai
2moA inteligência artificial do fim do mundo é um hype ou uma possibilidade? 👀 No episódio #7 do Segue o Fio, @sergio.nucleo.jor.br e @alexa…A inteligência artificial do fim do mundo é um hype ou uma possibilidade? 👀 No episódio #7 do Segue o Fio, @sergio.nucleo.jor.br e @…
↗Not briefed
arxiv
arxiv-ai
2moSharpness-Aware Minimization (SAM) Improves Classification Accuracy of Bacterial Raman Spectral Data Enabling Portable DiagnosticsAntimicrobial resistance is expected to claim 10 million lives per year by 2050, and resource-limited regions are most affected. Raman…
↗Not briefed
arxiv
arxiv-ai
2moGLAMDRING: Gait Learning And Morphology co-Design via Reinforcement LearnING of CPGsRobots are moving out of the structured factory floor and into unstructured environments such as disaster sites, planetary surfaces, a…
↗Not briefed
bluesky
bluesky-ai
2moDon't listen to meat-proxies People who just repeat what an llm saysDon't listen to meat-proxies People who just repeat what an llm says
↗Not briefed
bluesky
bluesky-ai
2moI'm having trouble caring about a lot of homebrew efforts unless I know it was made by someone actually interested in learning. So many …I'm having trouble caring about a lot of homebrew efforts unless I know it was made by someone actually interested in learning. So…
↗Not briefed
arxiv
arxiv-ai
2moThe syntax and semantics of goalsIn both cognitive science and computer science, goals are conceptualized as cognitive states that flexibly combine with world knowledg…
↗Not briefed
bluesky
bluesky-ai
2moAt least meat can be consumed and processed into safer forms. Meanwhile, the LLMs, not so much. We have diesel powered LLM datacenters in m…At least meat can be consumed and processed into safer forms. Meanwhile, the LLMs, not so much. We have diesel powered LLM datacenters…
↗Not briefed
bluesky
bluesky-ai
2mowhy choose boring planet-killing LLM-based AIs, when I'm right here ready to be your personal assistant? :3why choose boring planet-killing LLM-based AIs, when I'm right here ready to be your personal assistant? :3
↗Not briefed
arxiv
arxiv-ai
2moFrom Models to Systems: A Comprehensive Survey of Efficient Multimodal LearningThe rapid expansion of multimodal models has surfaced formidable bottlenecks in computation, memory, and deployment, catalyzing the ri…
↗Not briefed
arxiv
arxiv-ai
2moPredict Before You Deploy: Offline Prediction of Quantization-Induced Task Degradation for World Action ModelsWorld action models (WAMs) rely on video-generation backbones, requiring substantial memory and compute for deployment. Post-training…
↗Not briefed
hn
hn-ai
2moPrinciples for a New Utopianism—
↗Not briefed
bluesky
bluesky-ai
2mopeople could then choose from various “off the shelf” AI models. individuals would then fully own their personal model & host it privatel…people could then choose from various “off the shelf” AI models. individuals would then fully own their personal model & host it pri…
↗Not briefed
hn
hn-ai
2moCookie converter and Claude session check I use—
↗Not briefed
bluesky
bluesky-ai
2mo❣️ Google is right there brother I’m not an llm at your beck and call❣️ Google is right there brother I’m not an llm at your beck and call
↗Not briefed
bluesky
bluesky-ai
2moUsing AI to write your papers is not just wrong for all the obvious reasons; it also just breaks the entire model of academic journals, and…Using AI to write your papers is not just wrong for all the obvious reasons; it also just breaks the entire model of academic journals…
↗Not briefed
rss
marktechpost
2moKnowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens🔧 GPT-5GLiFormer Large scores 91.10 F1 on nested JSON, near GPT-5.6-luna's 91.96, while grounding every value in source spans. The post Knowl…
↗Not briefed