David Krueger
@DavidSKrueger
AI professor. Deep Learning, AI alignment, ethics, policy, & safety. Formerly Cambridge, Mila, Oxford, DeepMind, ElementAI, UK AISI. AI is a really big deal.
6/13 ICLR submissions accepted! (3 posters, 2 orals, 1 blog post) Congrats to @kayembruno Richard Turner @_clementneo @FazlBarez Neel Alex @ShoaibASiddiqui @_tom_bush @StephenLCasper @dhadfieldmenell and all the other authors! Summaries in thread below... 🧵🧵
This story is almost 100% on the risks and race dynamics; skepticism is an afterthought. The inexorable growth of AI x-risk awareness is truly heartening! It's not fast enough. And it's always 2 steps forward, 1 step back. But I've watched this trend for >15 years and it's…
This article from @TheEconomist offers an accurate overview of key dynamics shaping the development of AI today: the risks of the rapid race toward AGI and ASI, the challenges posed by open-sourcing frontier models, the deep uncertainty revealed by ongoing scientific debates and…
Huge new study finds: people think AI will worsen almost everything they care about. By a 3-to-1 margin, they want more regulation. And they deeply distrust the big AI labs. Feels like we’re approaching a social tipping point. How long until the backlash against Big AI begins?
I am so sick of people acting like positive thinking will save us from extinction. Does anyone actually belief this BS?
I've transcribed this further, because the clip below is just a tease. Keir Starmer: "I think AGI is going to be pretty amazing. And I'm a massive AI fan."
Another week, another member of Congress announcing their superintelligent AI timelines are 2028-2033:
if a sci-fi movie had AI safety researchers gravely alluding to “the MechaHitler Incident” it would seem so ridiculously overwrought
I think there are actually some deep underlying causes in common. These have to do with, e.g.: - lacking good coordination mechanisms - massive inequality corrupting democratic and epistemic institutions - development of tech for influencing human interactions and communities
It is passing strange that society seems to be going mad with hopelessness and despair, anger and hatred and sadism, loss of honor and kindness, a wanton destructiveness; and also the world is ending; but these two facts seem to be mostly unrelated.
Here’s what @Sama said about AI in 2015, before starting OpenAI: “WHY YOU SHOULD FEAR MACHINE INTELLIGENCE Development of superhuman machine intelligence (SMI) is probably the greatest threat to the continued existence of humanity. There are other threats that I think are more…
(1/2) Does it blow your mind that someone could be lying about lying down? It does for AI Overview.
🚀 Applications are open for the 2026 Horizon Fellowship! Deadline: Aug 28 Join a community of 80+ alums and spend up to two years in DC working on emerging tech policy at agencies, congress, or think tanks. Learn more and apply here: horizonpublicservice.org/applications-o…
Some personal news: Since leaving OpenAI, I’ve been writing publicly about how to build an AI future that’s actually exciting: avoiding the worst risks and building an actually good future. I’m excited to continue this work as a fellow of the Roots of Progress Institute.
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
This just in: OpenAI was the first ever AI lab to break the embargo on achieving gold medal performance on the International Mathematical Olympiad.
Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! deepmind.google/discover/blog/…
🚨 According to a friend, the IMO asked AI companies not to steal the spotlight from kids and to wait a week after the closing ceremony to announce results. OpenAI announced the results BEFORE the closing ceremony. According to a Coordinator on Problem 6, the one problem OpenAI…
1/ Are the safeguards in some of the most powerful AI models just skin deep? Our research on Jailbreak-Tuning reveals how any fine-tunable model can be turned into its "evil twin"—equally capable as the original but stripped of all safety measures.
AI industry lobbying + PACs will be the most well funded in history, making it all the more important to pass federal legislation soon before the process is completely corrupted