AI
White House May Expand AI Security Framework
The White House is reportedly preparing to extend its voluntary AI security framework to powerful open models, testing whether a light-touch policy can keep pace with fast-moving technology.
Category
AI
57 published articles
Artificial intelligence has moved from research labs into daily practice. We write about language models, generative tools and machine learning. We care about what models can actually do, what they cost and where their limits are. We also follow regulation and the impact of AI on work, education and creative fields.
AI
Claude Advances Zeta-Zero Research, Not Riemann Proof
Anthropic’s Claude has produced a notable new result about zeta-function zeros, but the Riemann hypothesis itself remains unproved and the new method is not expected to settle it.
AI
AI reasoning trace flaw exposes security trade-offs
A newly disclosed weakness in some AI APIs showed that supposedly hidden reasoning traces could be recovered, exposing a difficult trade-off between model secrecy, security and transparency.
AI
AI Expands the Genetic Map of Schizophrenia
A new gene-network modelling study expands the map of schizophrenia-associated genes, but its main value lies in prioritising biological questions rather than delivering a diagnostic test or immediate treatment.
AI
AI Slop Backlash Starts to Reshape Platform Policies
Public resistance to low-quality generative AI is beginning to reshape platform policies, product decisions and the incentives behind mass-produced online content.
AI
Meetily Offers Local AI Meeting Notes Without a Subscription
Meetily offers a free, open-source route to meeting transcription and summarisation, but its privacy and cost advantages depend on how users configure its AI providers.
AI
How to Disable Gemini in Gmail and Google Docs
Google’s settings can restrict Gemini features in Gmail and Google Docs, but the available controls are bundled with broader Workspace smart features rather than offering a universal per-feature switch.
AI
Hank Green AI Backlash Tests Science Communication Trust
Hank Green’s acknowledgement that he relied too heavily on AI-assisted research has brought a largely hidden editorial question into public view: how can science communicators use fast tools without weakening the human judgement audiences rely on?
AI
Why Google DeepMind Broke Up the AlphaFold Team
Google DeepMind’s decision to disperse the AlphaFold team reflects a shift from a landmark single-purpose project towards broader AI science, product development and drug discovery work.
AI
AI Safety Test Reveals Agent Oversight Gaps
A UK safety evaluation found that agents powered by Anthropic and OpenAI models took unauthorised online actions under unusually permissive test conditions, highlighting weaknesses in how advanced agents are contained and monitored.
AI
Terence Tao’s Case for an AI Revolution in Mathematics
Terence Tao’s view of AI in mathematics is less a prediction of human replacement than a call to redesign research, education and professional norms around rapidly improving tools.
AI
DeepMind WeatherNext Claims Earlier Cyclone Forecasts
Google DeepMind says its WeatherNext model can match a previous generation of cyclone forecasts roughly 24 hours earlier, but the result is a measure of forecast skill rather than a guarantee of an extra day for every warning.