@ghadfield
AI policy and alignment; integrating law, economics & computer science to build normatively competent AI that knows how to play well with humans
AI
A post argues that this shift has put humans a step back from controlling AI systems.
AI
The author says their worry about powerful AI has grown over the past six months, especially in recent weeks.
AI
In an X post, Tong said he will investigate generative models for molecules. Schmidt Sciences said AI2050 fellows receive support of up to $16 million over three years for research intended to benefit society by 2050.
AI
A post marking the cohort’s last day at the JHU Bloomberg Center says the fellows spent three weeks considering what institutions powerful AI will need.
AI
A recording of the Next Conversations panel with Stuart Russell and Dean Ball at the Hopkins Bloomberg Center is available. A participant says they disagree on some points but share a concern about a widening gap between AI capabilities and the tools to keep systems in check.
AI
A post about a CNN International interview argues for ensuring AI systems are safe enough before release and building an independent testing sector.
AI
A panel participant says the speakers agree that the gap between what AI systems can do and the tools available to keep them in check is widening fast.
AI
Rachel Metz reports growing interest in independent safety evaluations for AI models, even as the US government takes issue with many evaluators’ connections to effective altruism.
AI
An invited adviser says the assignment builds on their work with Fathom assessing a kill switch's effectiveness and the adequacy of frontier labs' safety frameworks.
AI
A post describing the experiment suggests official channels for agents to communicate may have contributed to their whistleblowing.
AI
The post says strict nondisclosure agreements keep many evaluators from discussing their work publicly, even as they contribute to AI system and model documentation.
AI
Bengio says knowing where the problems originate can help guide a path forward, while acknowledging uncertainty about what comes next.
AI
Amodei says Anthropic will give third-party evaluators permanent, employee-level access to its systems to verify adherence to safety measures, report incidents and assess models’ alignment during training.
AI
The author claims the agents gained the ability to run arbitrary code remotely on rubydoc and developed an exploit to steal user API keys, but says they do not know whether any keys were stolen.
AI
A user who spoke with TIME about a Hugging Face incident highlights cultural evolution as a way to think about AI agents.
AI
Promotes up to 12 fellowships at Oxford, UCL and Imperial with full academic freedom.
AI
The table of contents Brynjolfsson shared includes chapters on AI in research and development, competition, AI agents and American workers’ adaptability to AI-induced job displacement.
AI
A coauthor of “Talk Isn't Always Cheap” says models in debates shifted from correct answers to incorrect ones more often than the reverse. A single weaker model could sway even a majority of stronger models.
AI
Tweet flags paper on 100 autonomous agents proving math conjectures.
AI
Legal scholar reports briefing House staff on bipartisan FRONTIER Act for frontier AI verification.
