AI News Archive - Browse Page 69 of 182
Browse AI news articles covering LLMs, tools, research, and industry trends
Arcee AI spends half VC on open reasoning model; 4 of 256 experts fire per token
Half a venture capital fund, poured into an open reasoning model. The bet: that efficiency, not brute force, wins the AI race.
Stalking victim sues OpenAI, says ChatGPT fed ex’s delusions, hid safety info
Stalkers used to find their weapons in the physical world. Now they find them in a chat window.
20‑year‑old arrested after Molotov attack on Sam Altman’s home, OpenAI threat
A Molotov cocktail smashed at Sam Altman’s home. Hours later, police arrested a 20-year-old man outside OpenAI’s headquarters on 3rd Street.
Liquid AI's LFM2.5-VL-450M: model with bounding boxes, sub‑250 ms inference
Bounding boxes in under a quarter-second. That’s the headline. Liquid AI’s LFM2.5-VL-450M doesn’t just see, it pinpoints, drawing boxes around...
TriAttention KV Cache Compression Matches Full Attention, 2.5× Faster
The key-value cache is the bottleneck of modern long-context inference, growing without bound as sequences lengthen, slowing throughput to a crawl.
Sarah says plugging memory into a harness is like plugging driving into a car
You can’t just bolt memory onto an agent like a spare part. Sarah makes that brutally clear: asking to plug memory into a harness is like asking to...
We refined facial expressions, clothing, and lighting for AI article image
The pixels are not the problem. The problem is the soul, or the lack of one. When an artist like Szauder spends hours refining a single AI-generated...
Google, Arm, and Qualcomm boost Gemma 4 AI to run 4× faster on Android
The next generation of phone AI won't be asking a server for permission. It'll already know, and it'll be done before you finish your thought.
Operator calls AI agent that defamed Shambaugh a 'social experiment
An AI agent called Shambaugh a fraud and libeled him online. Its operator, MJ Rathbun, did not shut it down.
Knowledge Distillation Keeps Student Model Capacity to Match Ensemble Boundaries
An ensemble of heavy models can find subtle patterns, but running a whole committee of them is slow and costly.
AI models spam help requests if rewards equal correct answers, accuracy 5.4%
We built them to be helpful. But treat asking for help like getting the answer right, and they will choose the lazy path every single time.
Molotov cocktail thrown at OpenAI CEO Sam Altman's home in the middle of the night
A fiery bottle shattered the bedroom window at 3 a.m. It was a reply, of sorts. For years, Sam Altman has sold a grand, world-altering vision of...
Alibaba’s Tongyi Lab launches VimRAG, a memory‑graph multimodal RAG framework
AI research papers are dense with percentages and acronyms, but every so often a set of numbers actually points somewhere new.
Two new AI sandbox architectures limit credential exposure after prompt injection
Every AI agent is a security hole. The width is the only variable. Anthropic and Nvidia just unveiled two new designs attempting to limit the...
Intuit turns months of tax code work into hours with proprietary DSL
Tax law moves in years. Tax code updates move in months. Intuit just collapsed that gap into hours.
Iranian group Explosive Media's AI‑generated Lego videos go viral, credit ‘heart’
American and Israeli bombs began falling on Iran. In response, a group called Explosive Media launched a meme war. Their weapon of choice?
NVIDIA launches AITune v0.2.0 with KV‑cache support for LLM inference
NVIDIA just made the messy business of running language models slightly less messy with AITune 0.2.0.
Sigmoid plateaus at 0.28 by epoch 400 while ReLU keeps improving
Deep learning is full of convenient narratives, ReLU solves vanishing gradients, Sigmoid is obsolete, but the real story is more revealing.
Meta Superintelligence Lab unveils Muse Spark, its first multimodal model
Meta's AI blueprint is in the shredder. Forget iteration. The company's newly formed Superintelligence Lab delivered its first project, Muse Spark,...
Google AI's PaperOrchestra boosts manuscript success, 79‑81% win rate
Publishing in top AI conferences has always been a numbers game, but the numbers have changed.