Google DeepMind CEO Demis Hassabis warned that artificial general intelligence is years away and proposed a US-led industry-funded regulatory body modeled on FINRA, which critics say would entrench Big Tech incumbents by…
IBM warned that the artificial intelligence boom is squeezing enterprise software budgets, triggering a sector-wide rout in tech stocks. The caution from Big Blue raised questions about the sustainability of crypto-adjac…
Lumen Technologies launched a network threat intelligence integration with Palo Alto Networks, embedding Lumen Defender AMDR into Cortex XSIAM to accelerate triaging and reduce signal noise. The partnership aims to comba…
Anthropic released usage data from 1.2 million Claude Cowork sessions showing software development accounts for only 8.7% of usage, contrary to the product's marketing. Within 48 hours, OpenAI launched ChatGPT Work, a ne…
A developer demonstrates that fusing three small local language models (Llama 3 8B, Gemma 2 9B) at the logit level on a laptop can match the reasoning performance of Anthropic's Fable 5 multi-agent framework, challenging…
Tencent-backed Chinese startup StepFun launched the StepX Neo, claiming it as the world's first AI-native smartphone with an operating system designed for AI agents. The device features a personal AI assistant named Amoo…
Independent researchers Harry Brignull and Cennydd Bowles released Over The Edge 2.0, finding that Microsoft still uses deceptive design patterns in Windows, Edge, and Bing to coerce users into using Edge over alternativ…
Black Duck released AI-powered triage and CRA-ready compliance checks for its Coverity static analysis tool, marking the first integration of AI features into the two-decade-old product. The updates aim to address challe…
Elon Musk's AI company xAI installed 59 natural gas turbines for its Colossus 2 supercomputer project in Tennessee and Mississippi without obtaining required federal clean air permits, operating for months near predomina…
Tracebit researchers developed 'context bombs' that use defensive prompt injection to thwart AI-driven attacks by embedding decoy credentials and resources that trigger early warnings when targeted. The technique flips t…
Chinese AI startup DeepSeek is reportedly considering a new funding round at a $71 billion valuation, just two months after raising $7 billion at a $52 billion valuation. The company's valuation has tripled since April, …
Vilya-1, a new deep learning model for macrocyclic peptide design, improves geometric accuracy and chemical coverage over traditional methods, enabling generative applications for custom therapeutics. The model addresses…
AuditWeave, a new Python library, enhances accountability in AI-assisted decisions by creating a tamper-proof, hash-chained ledger that records each step in AI workflows. The tool, designed for regulated sectors like fin…
A new study on quantized language models reveals that 4-bit models offer an optimal balance between efficiency and reliability, showing unexpected robustness to real-world input perturbations. The findings challenge the …
A new study reveals that low-precision training can cause AI model parameters to freeze when gradient updates fall below half the unit in the last place (ULP) of a weight, effectively halting learning. The phenomenon was…
Ground truth datasets in machine learning are not objective but are constructed through human decisions and technological limitations, according to a new analysis. The piece argues that acknowledging the context-dependen…
Researchers propose a predictive divergence mask for reinforcement learning in large language models, offering potentially more stable updates than traditional trust-region methods like PPO. The approach uses closed-form…
Researchers introduced the Energy-guided Recursive Model (ERM), which uses Hopfield energies to improve AI reasoning, achieving high success rates on puzzles like Sudoku (98.97%) and mazes (99.30%). The model enhances sa…
Researchers propose a bilevel optimization framework to improve Direct Preference Optimization (DPO) for aligning large language models with human preferences, addressing noise in preference data. The method outperforms …
A Nobel laureate warns that society has a limited window to prepare for accelerating technological advancements, including AI and quantum computing, and that current efforts by governments, businesses, and educational in…