Twitxy Saved

Tim G. J. Rudner

Verified account This account is verified. Learn more

@timrudner

Assistant Professor, @UofT
CIFAR AI Chair, @VectorInst

AI Safety, Alignment, Scalable Oversight, Governance

Prev: Rhodes Scholar, @UniofOxford, @Yale
Toronto + Oxford + New Haven timrudner.com Joined December 2015
1K Following 3.2K Followers 1.8K Posts

Latest posts 5

Tim G. J. Rudner Verified account @timrudner
Advancing AI safety and AI alignment is more important than ever.

Apply to the @aims_oxford PhD program at the @UniofOxford if you want to join my group and work with me on AI safety and alignment!

🌐More details: timrudner.com/joining
Fully Funded 4-year Doctoral Studentship Joint with Oxford Martin AI Governance Initiative and the EPSRC CDT in Autonomous Intelligent Machines & Systems (AIMS) October 2027 start. Further details here: aims.robots.ox.ac.uk
View quoted post on X
0 replies 3 reposts 59 likes 4.3K views View on X
Tim G. J. Rudner Verified account @timrudner
What a great conversation -- highly recommend!

If you want to learn more about why AI agents may go rogue and become 'misaligned', @hlntnr and I wrote a non-technical explainer paper on this back in 2021:

cset.georgetown.edu/publication/ke…

It's more relevant today than ever!
cset.georgetown.edu Key Concepts in AI Safety: Specification in Machine Learning | Center for Security and Emerging... This paper is the fourth installment in a series on “AI safety,” an area of machine learning research that aims to identify causes of unintended behavior in machine learning systems and develop tools...
Jon Favreau Verified account @jonfavs Sep 19
Why can't we just unplug the AI?

Was the Hugging Face attack really that bad?

Why do the AI companies want to put AI models in charge of building new AI models?

AI researcher and former @OpenAI board member @hlntnr helped me answer all these Qs and more on today's Offline
View quoted post on X
0 replies 0 reposts 7 likes 1.2K views View on X
Tim G. J. Rudner Verified account @timrudner
OpenAI was hacked by three people using Claude Opus 5.

It doesn't take much creativity to imagine what kind of damage near-future rogue agents would be able to cause as capabilities keep improving---and we need better scalable oversight and verification methods to prevent that.
s1r1us Verified account @S1r1u5_ Sep 18
On July 25, we hacked OpenAI.

Two bugs let us take over ChatGPT/Codex accounts of OpenAI employees (+some unaffiliated users) and reach connected services: Outlook, Slack, GitHub, etc.

We proved it with a PR in OpenAI’s internal codebase . It took us <72h. 🧵
View quoted post on X