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
2mo“An artificial intelligence researcher at Anthropic publicly quit his position 15/9 warning that major AI companies care too much about win…“An artificial intelligence researcher at Anthropic publicly quit his position 15/9 warning that major AI companies care too much abou…
↗Not briefed
bluesky
bluesky-ai
2moBecause they’re also pretending they can impartially regulate each other www.verysane.ai/p/is-metr-a-...Because they’re also pretending they can impartially regulate each other www.verysane.ai/p/is-metr-a-...
↗Not briefed
bluesky
bluesky-ai
2moOh there’s a fun story hook. Civilization receives 2+ LLM probes at roughly the same time, we’ve got a CCP one, we’ve got a Claude, we’ve g…Oh there’s a fun story hook. Civilization receives 2+ LLM probes at roughly the same time, we’ve got a CCP one, we’ve got a Claude, we…
↗Not briefed
bluesky
bluesky-ai
2moIf you run an LLM and ask it to do something, it is your PC acting as the harness - via Claude Code or a web browser A web browser is a sa…If you run an LLM and ask it to do something, it is your PC acting as the harness - via Claude Code or a web browser A web browser is…
↗Not briefed
hn
hn-ai
2moSoftware Citations from Replication Files—
↗Not briefed
arxiv
arxiv-ai
2moA Calibrated Instrument for Measuring How Inference Optimizations Affect Output QualityLarge language model optimization is an active research area, spanning quantization of model weights, early-exit methods for skipping…
↗Not briefed
hn
hn-ai
2moOne seeded bug, 26 AI agents: all passed the tests, all stayed broken—
↗Not briefed
bluesky
bluesky-ai
2moi actually did write about this before but honestly i only did it bc i wrote the best lede of all time here techcrunch.com/2023/11/21/w...i actually did write about this before but honestly i only did it bc i wrote the best lede of all time here techcrunch.com/2023/11/21/…
↗Not briefed
arxiv
arxiv-ai
2moMissing Bridges: Composition-Aware Active Imitation LearningActive imitation learning reduces expert effort by allowing a learner to request the demonstrations it needs. Existing methods typical…
↗Not briefed
bluesky
bluesky-ai
2moIdk. I go back and forth on this. I readily say things like “OpenAI is building hackerbots that are autonomously carrying out cyberattacks …Idk. I go back and forth on this. I readily say things like “OpenAI is building hackerbots that are autonomously carrying out cyberatt…
↗Not briefed
bluesky
bluesky-ai
2moGood to see there’s idiots in Canada as wellGood to see there’s idiots in Canada as well
↗Not briefed
bluesky
bluesky-ai
2mospineless shits "deputy prime minister Richard Marles, said the federal government was “working at a pace” to respond to the issue of copy…spineless shits "deputy prime minister Richard Marles, said the federal government was “working at a pace” to respond to the issue of…
↗Not briefed
bluesky
bluesky-ai
2moThe AI doom panic is a farce. OpenAI has been irresponsibly training their agents to hack other companies which can then allow these same m…The AI doom panic is a farce. OpenAI has been irresponsibly training their agents to hack other companies which can then allow these s…
↗Not briefed
hn
hn-ai
2moShow HN: Review agent changes locally before pushingI've often found that agents do stupid things when implementing requests. When I only see their changes with limited context in…
↗Not briefed
bluesky
bluesky-ai
2moLLM helps weird, amoral dude design Plague Inc winner with home CRISPR kit feels like the main threat.LLM helps weird, amoral dude design Plague Inc winner with home CRISPR kit feels like the main threat.
↗Not briefed
bluesky
bluesky-ai
2moapropos of EA/LLM discourse, I hypothesize that the creators and early advocates of a given technology are rarely correct about its longer …apropos of EA/LLM discourse, I hypothesize that the creators and early advocates of a given technology are rarely correct about its lo…
↗Not briefed
arxiv
arxiv-ai
2moThe Attention Within: Consensus Dynamics in Selective State Space Models🔧 ConsensusSelective state space models (SSMs) have recently emerged as a compelling alternative to transformers, combining competitive performan…
↗Not briefed
arxiv
arxiv-ai
2moModeling the Developmental Shift in Telicity AcquisitionAcquiring telicity, which is the distinction between bounded (e.g., ate an apple) and unbounded (e.g., ate apples) events, requires fi…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
rss
bing-news-ai
2moArtificial Intelligence (AI) May Have a New Bear Thesis: Human ExtinctionAn artificial intelligence researcher who worked at OpenAI and Anthropic recently warned that there is a real chance that AI could wip…
↗Not briefed
arxiv
arxiv-ai
2moQuanText: Protecting Dataset-Level Secrets in Textual Data SharingNatural-language datasets support many downstream applications and research studies, but releasing text can reveal sensitive global pr…
↗Not briefed
hn
hn-ai
2moGrandma disables a fiber optic drone—
↗Not briefed
bluesky
bluesky-ai
2moMicrosoft agrees to new AI safety rules for US schoolsMicrosoft agrees to new AI safety rules for US schools
↗Not briefed
bluesky
bluesky-ai
2moit doesn't need to be sentient, but what goes on within an llm is complex enough to be hard to predict. the llm can seem to understand and …it doesn't need to be sentient, but what goes on within an llm is complex enough to be hard to predict. the llm can seem to understand…
↗Not briefed
bluesky
bluesky-ai
2moNEW: Senior AI policy advisor at White House OSTP Sihao Huang, who has been deeply involved in developing frontier AI policy, is leaving to…NEW: Senior AI policy advisor at White House OSTP Sihao Huang, who has been deeply involved in developing frontier AI policy, is leavi…
↗Not briefed
bluesky
bluesky-ai
2moEXCLUSIVE: Senior Trump administration officials met with Anthropic’s top Washington executive to discuss artificial intelligence safety ri…EXCLUSIVE: Senior Trump administration officials met with Anthropic’s top Washington executive to discuss artificial intelligence safe…
↗Not briefed
hn
hn-ai
2moShow HN: Thurbox – A tmux-based TUI and CLI for local AI agent orchestration—
↗Not briefed
arxiv
arxiv-ai
2moThe Operable Pareto Front: Distilling Offline Search into Run-Time Control for Multi-Objective UAV Edge-Computing SchedulingA UAV mobile edge computing (MEC) fleet trades energy against delay, and its schedules form a Pareto front; we call a scheduler operab…
↗Not briefed
bluesky
bluesky-ai
2moSame energy as all the tech bros currently building coder/reviewer LLM loops and feeling like geniuses because they built baby's first ML o…Same energy as all the tech bros currently building coder/reviewer LLM loops and feeling like geniuses because they built baby's first…
↗Not briefed
arxiv
arxiv-ai
2moFourier Analysis of Parametrized Interactive Quantum ClassifiersInteractive Quantum Classifiers (IQCs) constitute a family of quantum machine learning models inspired by open quantum systems, in whi…
↗Not briefed
arxiv
arxiv-ai
2moWhom Do AI Agents Work For? Role Assignment Induces Sponsorship Bias in LLM Recommenders🔧 ReplicateLarge language models (LLMs) now serve as conversational shopping assistants on platforms that also sell advertising. These AI agents…
↗Not briefed
bluesky
bluesky-ai
2moDo you really want to suggest that the LLM is reasoning in a state that lacks conscious control?Do you really want to suggest that the LLM is reasoning in a state that lacks conscious control?
↗Not briefed
rss
techmeme
2moAnalysis: only ~2.3 GW of planned data center capacity is delayed by direct project-level policies like local moratoriums and New York's data center EO (SemiAnalysis)SemiAnalysis: Analysis: only ~2.3 GW of planned data center capacity is delayed by direct project-level policies like local moratorium…
↗Not briefed
bluesky
bluesky-ai
2mothis is one of the few times i would recommend you get an LLM to do itthis is one of the few times i would recommend you get an LLM to do it
↗Not briefed
arxiv
arxiv-ai
2moMultimodal Conditioning of Fine-Tuned Stable Diffusion XL for Controllable and Culturally Faithful Ulos Motif Generation🔧 Stable DiffusionThe traditional Batak Ulos weaving industry faces growing challenges in producing diverse, innovative motifs due to limitations in con…
↗Not briefed
arxiv
arxiv-ai
2moRideWay: Benchmarking Efficient Task Completion for Tool-Using Language AgentsAI agents are usually evaluated by whether they complete a task. In interactive service settings, a successful agent can still frustra…
↗Not briefed
bluesky
bluesky-ai
2moIf an LLM logically deduces the cure to cancer from first principles and supplies a 600k line Lean proof establishing as much I will be the…If an LLM logically deduces the cure to cancer from first principles and supplies a 600k line Lean proof establishing as much I will b…
↗Not briefed
bluesky
bluesky-ai
2mo!!!!!!! www.infosecurity-magazine.com/news/attacke...!!!!!!! www.infosecurity-magazine.com/news/attacke...
↗Not briefed
arxiv
arxiv-ai
2moTuiML: Machine Learning for AI AgentsMachine-learning libraries such as Weka and scikit-learn were designed for human programmers. Language-model agents now use these same…
↗Not briefed
arxiv
arxiv-ai
2moContiguity, Not Importance: Budgeted Repair of Stale KV Caches After Document EditsKV-cache reuse can reduce inference cost in retrieval-augmented generation and agentic systems, but cached contexts may become stale w…
↗Not briefed
bluesky
bluesky-ai
2moIt gets sillier than that though. Let's say the LLM doesn't have access to a record of its chain of thought. Somebody's going to have troub…It gets sillier than that though. Let's say the LLM doesn't have access to a record of its chain of thought. Somebody's going to have…
↗Not briefed
bluesky
bluesky-ai
2moLots of folks say OpenAI's sandbox was negligent but (a) the agents used *several* 0day to break out of the sandbox (b) find me literally a…Lots of folks say OpenAI's sandbox was negligent but (a) the agents used *several* 0day to break out of the sandbox (b) find me litera…
↗Not briefed
arxiv
arxiv-ai
2moEncoder Awakening via Adapters: Effective Domain-Adaptive Fine-tuning of Speech-LLMsSpeech Large Language Models (Speech-LLMs), typically built from a pre-trained speech encoder, a modality projector, and an LLM fine-t…
↗Not briefed
bluesky
bluesky-ai
2moIt is already killing people. Palantir + Claude gave tRump/Hegseth that school of girls in Iran as a target. It doesn't need to become supe…It is already killing people. Palantir + Claude gave tRump/Hegseth that school of girls in Iran as a target. It doesn't need to become…
↗Not briefed