💡Have an idea for using #AI to enhance teaching and learning? We're offering up to £10,000 to support practical, educator-led projects exploring AI in education at @cam.ac.uk 📅 Applications close 24 September ➡️ www.ai.cam.ac.uk/calls/ai-for...

This is so frustrating to read from The New Yorker. Reward hacking *long* predates LLMs. The Coast Runners paper on reward hacking (in reinforcement learning, pre-LLMs) is 10 years old! Reward hacking is not new, and not a sign of a breakthrough intelligence, much less sentience. Did AI write this?

An important message. The phenomenon is a repeat pattern for UK … I worry that it’s because there’s a sensitive breed of superficial ‘experts’ that doesn’t like it when their confident pronouncements are undermined by data and evidence. (feel free to undermine this idea with data and evidence)

Royal Statistical Society @royalstatsoc.bsky.social · 5d ago

The chair of our AI Task Force has an opinion piece in @civilserviceworld.bsky.social arguing that separating the parts of government dealing with AI policy from digital and data policy risks making the government's AI strategy harder to deliver. www.civilserviceworld.com/professions/...

How online betting is destroying young people's lives in South Africa | DW Reporter www.youtube.com/watch?v=-13y...

How online betting is destroying young people's lives in South Africa | DW Reporter

Sisanda Lwana lost over 7000 Euros betting online - and overcame her addiction. Her story brings to light the dark side of the gambling industry. Artikeltext The dream of getting rich…

youtube.com

The challenge of the "social productivity paradox" in AI and our ideas about how to address it with an "attention reinvestment cycle" are now available in Science and Public Policy. academic.oup.com/spp/advance-...

Mind the gap: connecting AI innovation to widespread public value

Abstract. Public dialogues have produced clear demand signals for AI. These dialogues imagine innovations that improve our shared wellbeing and prosperity

academic.oup.com

New report just out from the @ai.cam.ac.uk team exploring what the public wants from AI in local government. ➡️ Check out the thread below and the full report for further insights.

ai@cam@ai.cam.ac.uk · 2w ago

NEW REPORT 🚨📚 What does the public want from #AI in local government? Our new report draws on public dialogue with Cambridgeshire residents. The conversation is changing. The question is no longer whether to use AI, but where it can make a real difference. ➡️ www.ai.cam.ac.uk/reports/publ...

I've really enjoyed this public dialogue we convened in collaboration with Hopkins van Mil. Here's our full report on what the public think about the use of AI in Local Government.

ai@cam@ai.cam.ac.uk · 2w ago

NEW REPORT 🚨📚 What does the public want from #AI in local government? Our new report draws on public dialogue with Cambridgeshire residents. The conversation is changing. The question is no longer whether to use AI, but where it can make a real difference. ➡️ www.ai.cam.ac.uk/reports/publ...

Did @cbsnews.com.web.brid.gy with Major Garrett last night on the OpenAI/Hugging Face cybersecurity event. Interview available here www.youtube.com/watch?v=bH9A...

Machine learning professor breaks down OpenAI model's hack of another AI company

YouTube video by CBS News

youtube.com

Spent a day being a media talking head on the OpenAI / @hf.co hack. It's another example of how the way we think about #AISafety is failing. It's idealised, and doesn't fit how companies operate. More details in my recent Trent AI blog here: trent.ai/blog/judgmen...

The Judgment Layer: Rethinking AI Safety for Agentic Systems | Trent AI

As AI agents become more autonomous, safety depends on preserving human judgment. Learn why the judgment layer is essential for secure agentic AI.

trent.ai

OpenAI's latest #AI security incident has sparked renewed debate about AI safety. Our Chair of ai@cam, @lawrennd.bsky.social, spoke to #BBCNews about what this reveals about today's AI models & why commercial pressure must not come at the expense of safe deployment. www.bbc.co.uk/news/article...

OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack

It is one of the first publicly disclosed cyber-attacks carried out by AI without direct human involvement.

bbc.co.uk

Great thread to summarise

Weeks ago, @chanda.blacksky.app sent me this declaration, written and signed by a bunch of mathematicians. But I only had time to read it yesterday, which ended up being right on time given a new round of hype falling exactly under the patterns this document notes.🧵 leidendeclaration.ai

Leiden Declaration on Artificial Intelligence and Mathematics

This declaration calls for action to address the challenges posed by the use of artificial intelligence within mathematics research.

leidendeclaration.ai

‼️Don't miss your chance to join our AI-deas Sprint FAQ session tomorrow! If you have an #AI research idea & considering applying for up to £25,000, this is your opportunity to ask questions & find out more. 🕤 9.30am | Wed 22 July 📍 Zoom ➡️ Register here cam-ac-uk.zoom.us/meeting/regi...

Volume 326 Proceedings of Second Workshop on Geometry-grounded Representation Learning and Generative Modeling proceedings.mlr.press/v326/ Is now available on PMLR

Volume 323 proceedings.mlr.press/v323 Proceedings of Causal Learning and Reasoning 2026 Is now available on PMLR.

Huge thanks to @lawrennd.bsky.social for a brilliant and very well attended talk last night. We're back in Sept. with @variolator.bsky.social who'll be telling us about 'The Common Cold: 30,000 Years of the Sniffles' - sitp.online/sitp/bedford/

How should teaching & learning evolve in a world with AI? We're offering up to £10,000 to test new teaching approaches with #AI. Open to academics, examiners, education managers & course administrators - anyone who shapes the teaching experience! Find out more ⬇️ www.ai.cam.ac.uk/calls/ai-for...

Prof @dianecoyle1859.bsky.social argues in her essay that the arms race framing not only distorts politics, but also actively undermines economic security.

We're back this week. Come along on Thurs. 16th July (no football that night) when @lawrennd.bsky.social will explain 'What an I Can Do That A.I. Can’t'. sitp.online/sitp/bedford/

AI regulation needs statistics. Without robust evaluation of performance, fairness and uncertainty, the most important questions about AI remain unanswered. Read our new report 👉 rss.org.uk/news-publica... Browse the gallery below to explore our case studies and recommendations 👇

As always, thanks to @bedfordindependent.co.uk for including next week's talk with @lawrennd.bsky.social in their indispensable What's On guide.

Bedford Independent@bedfordindependent.co.uk · 4w ago

It's what's on Wednesday, and we're gearing ourselves up for the best weekend of the year - Bedford River Festival. Don't worry, we've also got the best of the Borough's live music, theatre, comedy, guided walks, art, culture and more. Kindly sponsored by Bedford Esquires #Bedford #BedfordNews

Small nutshell here; both reports mentioned are online with summaries www.project-syndicate.org/commentary/c...

Can the Manchester Model Save Britain?

Diane Coyle considers some of the lessons Andy Burnham should take with him to Downing Street.

project-syndicate.org

Some commentary from me on Andy Burnham's big economics speech, focusing on his theory of local economic growth, the role of universities, innovation and skills. With some history of what's behind this policy agenda softmachines.org?p=3324

Good analysis from @richardaljones.bsky.social on @andyburnham.bsky.social’s economic and innovation policy agenda including this lovely reference to @dianecoyle1859.bsky.social softmachines.org?p=3324

Volume 333 Proceedings of Conference on Health, Inference, and Learning 2026 proceedings.mlr.press/v333/ Is now available on PMLR

We're back this month. Come along on Thurs. 16th July when @lawrennd.bsky.social will explain 'What an I Can Do That A.I. Can’t'. sitp.online/sitp/bedford/ Nothing artificial about this talk, folks. #AI

Volume 319 Proceedings of IndabaX Nigeria 2026 proceedings.mlr.press/v319 are now available on PMLR @deeplearningindaba.bsky.social

Volume 325 Proceedings of Geometry, Topology, and Machine Learning 2025 proceedings.mlr.press/v325/ Is now available on PMLR.

🚀 Round two of the AI-deas Sprint Programme is now open for applications! 🚀 Got an #AI research idea with real-world potential? We're offering Cambridge researchers up to £25,000 to build a working prototype. Find out more and apply here ➡️ www.ai.cam.ac.uk/calls/ai-dea...

Volume 331 Proceedings of Learning for Dynamics and Control 2026 proceedings.mlr.press/v331/ Is now available on PMLR.