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 researcher Daniel Selsam confirming the above. "even knowing all the mistakes that were made, one would not have predicted that the …OpenAI researcher Daniel Selsam confirming the above. "even knowing all the mistakes that were made, one would not have predicted that…
↗Not briefed
bluesky
bluesky-ai
2moWhat's their incentive, you might ask: simple. You'd only need a few such incidents for security people to start taking independent LLM swa…What's their incentive, you might ask: simple. You'd only need a few such incidents for security people to start taking independent LL…
↗Not briefed
arxiv
arxiv-ai
2moBalancing Trial and Reorder: A Hybrid Sequential Transformer-GBDT Ranker for On-Demand DeliveryOn a delivery platform, personalized store ranking greatly influences what users find and order. Unlike digital-only domains, candidat…
↗Not briefed
bluesky
bluesky-ai
2moI've replaced a few services with stuff I've built with Claude... More on the horizon but I cringe when I see people say I've never coded a…I've replaced a few services with stuff I've built with Claude... More on the horizon but I cringe when I see people say I've never co…
↗Not briefed
hn
hn-ai
2moIs GLM-5.3-Flash Mythos-Level at Cyber?—
↗Not briefed
bluesky
bluesky-ai
2moAI safety people are always like "what if they hack the nukes" instead of "what if self-propagating LLM script kiddie swarms hack the worst…AI safety people are always like "what if they hack the nukes" instead of "what if self-propagating LLM script kiddie swarms hack the…
↗Not briefed
bluesky
bluesky-ai
2moAcho que foi desde q o Xi Jinping fez um pronunciamento em julho sobre modelos de código aberto. Os investidores lá da China estavam espera…Acho que foi desde q o Xi Jinping fez um pronunciamento em julho sobre modelos de código aberto. Os investidores lá da China estavam e…
↗Not briefed
arxiv
arxiv-ai
2moPhysics Informed Random Feature Neural Networks for Solving PDEsMachine learning-based partial differential equations (PDEs) solvers have attracted significant attention in recent years. Most progre…
↗Not briefed
bluesky
bluesky-ai
2mo「AI配管工のスーパーダリオ」で笑ってしまった😂 (ダリオはAnthropic社のCEOのファーストネーム)「AI配管工のスーパーダリオ」で笑ってしまった😂 (ダリオはAnthropic社のCEOのファーストネーム)
↗Not briefed
bluesky
bluesky-ai
2moYes - do I really need to pay a bunch of money for someone else’s SAAS solution if I at least THINK I can roll my own with a LLM that’s con…Yes - do I really need to pay a bunch of money for someone else’s SAAS solution if I at least THINK I can roll my own with a LLM that’…
↗Not briefed
bluesky
bluesky-ai
2moThis is the correct answer to OpenAI and any other greedy, thieving AI companies executing their playbook But of course, Albanese, Bowen, …This is the correct answer to OpenAI and any other greedy, thieving AI companies executing their playbook But of course, Albanese, Bo…
↗Not briefed
arxiv
arxiv-ai
2moImplementing a White-Box Undetectable Backdoor for Random Fourier FeaturesGoldwasser et al. showed that undetectable backdoors can be planted in machine learning models trained with the Random Fourier Feature…
↗Not briefed
bluesky
bluesky-ai
2moI don’t know who needs to hear this but please don’t plug shit other people have made into a LLM. That’s not cool or fun for the creator to…I don’t know who needs to hear this but please don’t plug shit other people have made into a LLM. That’s not cool or fun for the creat…
↗Not briefed
bluesky
bluesky-ai
2moHey, remember this for when AI kills us all: Trump says he alone is responsible for keeping it in check. I'm gonna say "I told you so" wit…Hey, remember this for when AI kills us all: Trump says he alone is responsible for keeping it in check. I'm gonna say "I told you so…
↗Not briefed
bluesky
bluesky-ai
2moJensen knows that any "slowdown" is fucking Ragnarok for the AI bubble he's currently getting rich off of. www.wheresyoured.at/ai-is-alread…Jensen knows that any "slowdown" is fucking Ragnarok for the AI bubble he's currently getting rich off of. www.wheresyoured.at/ai-is-a…
↗Not briefed
bluesky
bluesky-ai
2moIf a company forces me to chat with an LLM before I can access a human representative for assistance, I will seek to replace their services…If a company forces me to chat with an LLM before I can access a human representative for assistance, I will seek to replace their ser…
↗Not briefed
hn
hn-ai
2moShow HN: Nowdex – AI agent usage on your iPhoneHi HN,<p>I&#x27;ve been using Claude Code, Codex and Cursor quite a bit lately, and I found myself checking their usage limits all the…
↗Not briefed
bluesky
bluesky-ai
2moI'll be more bullish on Ai when the following things happen: 1. Model Collapse is solved, if not among open weight models (the Chinese one…I'll be more bullish on Ai when the following things happen: 1. Model Collapse is solved, if not among open weight models (the Chines…
↗Not briefed
arxiv
arxiv-ai
2moNegation Beyond the Verbal Channel: Temporal Multimodal Correlates in DialogueNegation is typically modeled through its linguistic realization, although spoken interaction is accompanied by tightly coordinated no…
↗Not briefed
bluesky
bluesky-ai
2moOr maybe I could just ask an LLM to summarise the books for me, that would no doubt be equally fun and considerably more efficientOr maybe I could just ask an LLM to summarise the books for me, that would no doubt be equally fun and considerably more efficient
↗Not briefed
bluesky
bluesky-ai
2moLol this guy hypes the stock market every time he opens his yap cos it's the only thing that hasn't yet gone to shit in his second term. …Lol this guy hypes the stock market every time he opens his yap cos it's the only thing that hasn't yet gone to shit in his second ter…
↗Not briefed
hn
hn-ai
2moCatching structural lookahead bias in Python backtests—
↗Not briefed
bluesky
bluesky-ai
2moAn OpenAI model broke containment & hacked Hugging Face to cheat a benchmark. #GIEE is asking: how do we evaluate agentic AI we can't fully…🔧 Hugging FaceAn OpenAI model broke containment & hacked Hugging Face to cheat a benchmark. #GIEE is asking: how do we evaluate agentic AI we can't…
↗Not briefed
arxiv
arxiv-ai
2moParsHate: A Benchmark Dataset for Hate and Target Detection in PersianWe introduce ParsHate, a manually annotated dataset of 10,000 Persian tweets spanning 2013-2022, representing the first decade-long be…
↗Not briefed
bluesky
bluesky-ai
2molook, i know everyone wants these things to be smoke and mirrors, but this is just human-feedback reinforcement learning. this is "click al…look, i know everyone wants these things to be smoke and mirrors, but this is just human-feedback reinforcement learning. this is "cli…
↗Not briefed
arxiv
arxiv-ai
2moWhere Post-Training Quantization Breaks Text Embedders: A Measured Map Across Four Embedder FamiliesWeight-only post-training quantization is the cheapest way to shrink a retrieval embedder, and the received advice for applying it --…
↗Not briefed
bluesky
bluesky-ai
2moAnthropic is! (But if this is just accounting smoke & mirrors I would take a loose analogy to banking, which also benefits from Monopoly pr…Anthropic is! (But if this is just accounting smoke & mirrors I would take a loose analogy to banking, which also benefits from Monopo…
↗Not briefed
hn
hn-ai
2moA zero-dependency Python loop detector for AI agents—
↗Not briefed
rss
marktechpost
2moAgent Harness vs Agent Framework vs MCP: Which Layer Owns the Loop, State, Tools, Permissions, and RecoveryA practitioner's map of the 3 layers in a modern agent stack, with verified sources and an overlap analysis. The post Agent Harness vs…
↗Not briefed
bluesky
bluesky-ai
2moMost people who haven’t cooked their brains working in Ant or OpenAI feel the same way, at least about transformers. I’ve been wrong about …Most people who haven’t cooked their brains working in Ant or OpenAI feel the same way, at least about transformers. I’ve been wrong a…
↗Not briefed
bluesky
bluesky-ai
2moI'd say this is a huge security breach, but at this point...I'd say this is a huge security breach, but at this point...
↗Not briefed
rss
techcrunch-ai
2moNvidia CEO Jensen Huang tells Trump ‘we’re not going to let [an AI slowdown] happen’Though Elon Musk and Sam Altman have supported Dario Amodei's calls to slow the pace of AI development, Jensen Huang seems to feel dif…
↗Not briefed
arxiv
arxiv-ai
2moAttention Mean Fields Predict Average Representation Dynamics and Reveal Context-Specific ComputationA language model's representation geometry is not predetermined; it evolves as the model runs. A faithful account of that geometry mus…
↗Not briefed
hn
hn-ai
2moA Letter from a Machine Learning Engineer—
↗Not briefed
bluesky
bluesky-ai
2moOpenAI is well known for hiring only 'true AGI believers' as developers. So ok to ignore what they say. Personally I think the AI-as-a-hac…OpenAI is well known for hiring only 'true AGI believers' as developers. So ok to ignore what they say. Personally I think the AI-as-…
↗Not briefed
hn
hn-ai
2moAI Doomer Hypeloop Suspiciously Serves to Support OpenAI and Anthropic's Agenda—
↗Not briefed
hn
hn-ai
2moWant to stop AI from destroying the world? Make it Open SourceMake Claude and Chat GPT open source. Give everyone the ability to build advanced AI systems. Don&#x27;t take all the guns, give the…
↗Not briefed
bluesky
bluesky-ai
2moI really hate the fact that it would've cost a lot less if the AI ram crisis wasn't happening, I don't think I'm upset enough at the fact t…I really hate the fact that it would've cost a lot less if the AI ram crisis wasn't happening, I don't think I'm upset enough at the f…
↗Not briefed
bluesky
bluesky-ai
2motheir AI is “too dangerous to advance quickly” turns failure into theatre, buying time for a ground‑up paradigm reboot. #Anthropic #OpenAI …their AI is “too dangerous to advance quickly” turns failure into theatre, buying time for a ground‑up paradigm reboot. #Anthropic #Op…
↗Not briefed
bluesky
bluesky-ai
2moCalls for an #AI slowdown from Altman, Amodei, and Musk now unexpectedly in lockstep aren’t about existential fear. They’re about the bruta…Calls for an #AI slowdown from Altman, Amodei, and Musk now unexpectedly in lockstep aren’t about existential fear. They’re about the…
↗Not briefed
arxiv
arxiv-ai
2moBounded Adjustment with Reliability-Guided Embedding for Imbalanced Learning with Noisy LabelsClass-balanced learning and label noise create a coupled failure mode: frequency correction prevents majority classes from dominating…
↗Not briefed
bluesky
bluesky-ai
2moHoly sh** WTH? www.navytimes.com/news/your-na...Holy sh** WTH? www.navytimes.com/news/your-na...
↗Not briefed
bluesky
bluesky-ai
2mothe only LLM i can endorse is a Large Lego Modelthe only LLM i can endorse is a Large Lego Model
↗Not briefed
bluesky
bluesky-ai
2mothis is what an acausal trade looks like to an LLMthis is what an acausal trade looks like to an LLM
↗Not briefed
hn
hn-ai
2moAccount Deactivated Due to Biological Research—
↗Not briefed
arxiv
arxiv-ai
2moCertified Uncertainty Propagation in One-Shot Federated Bayesian Models via Posterior Event TransportProbabilistic certification of Bayesian neural networks lower-bounds the posterior probability that a model satisfies a verifier-defin…
↗Not briefed
arxiv
arxiv-ai
2moRegister Tokens for Bounded-State Reasoning in Diffusion Language ModelsMasked diffusion language models (dLLMs) generate text by iteratively denoising masked tokens with bidirectional attention. Extending…
↗Not briefed
bluesky
bluesky-ai
2moLast week a former Anthropic safety researcher publicly expressed his concerns about AI, arguing that there was a 10% chance the emerging t…Last week a former Anthropic safety researcher publicly expressed his concerns about AI, arguing that there was a 10% chance the emerg…
↗Not briefed
hn
hn-ai
2moShow HN: VideoHighlighter – offline self-hosted video analyzerHey HN,<p>I created this project starting with subtitles creation for my 7 years old son. Than added video highlighting and finally it…
↗Not briefed
arxiv
arxiv-ai
2moFast-Convergent Meta-RL via Gradient-Clustered BS Sampling for Edge CachingWireless edge caching networks typically consist of many independent Base Stations (BSs), each facing its own request rate and content…
↗Not briefed