@ChrisPainterYup Also interesting to see the grouping by lab!
English
Jeremie Harris
2.4K posts
Jeremie Harris
@jeremiecharris
Co-founder & CEO of Gladstone AI We promote responsible AI R&D and adoption by designing and deploying safeguards against AI-driven national security threats.
Mountain View, CA Katılım Kasım 2017
755 Takip Edilen5.8K Takipçiler
It’s kind of a surprising fact that humanity is training models with enough intellectual diversity on this axis that this kind of question is actually a scissor for them
Jan Kulveit@jankulveit
Asked AIs the Red/Blue button question. Lots to notice, but posting without further commentary. First plot is with max reasoning, models called via API.
English
Jeremie Harris retweetledi
That’s a stretch, @nytimes!
The Action Plan calls on the U.S. to accelerate AI innovation while simultaneously investing in AI interpretability and biosecurity, evaluating national security risks in frontier models, and combating synthetic media.
English
Jeremie Harris retweetledi
TODAY: TRUMP’S A.I. ACTION PLAN
The president is set to sign executive orders that will solidify the U.S. as the global AI leader.
4PM ET: @jeremiecharris & @harris_edouard join to break down what this means.
English
This is a great document, right on the money on key national security issues, and striking the right balance. Reads like the work of builders with their eye on the ball.
Looking forward to discussing it on @WillCainShow_ at 4pm ET
Director Michael Kratsios@mkratsios47
Today the @WhiteHouse released America’s AI Action Plan to win the global race. We need to OUT-INNOVATE our competitors, BUILD AI & energy infrastructure, & EXPORT American AI around the world. Visit AI.gov
English
Jeremie Harris retweetledi
Today the @WhiteHouse released America’s AI Action Plan to win the global race.
We need to OUT-INNOVATE our competitors, BUILD AI & energy infrastructure, & EXPORT American AI around the world.
Visit AI.gov
English
@ChrisPainterYup How do you see the synthesis between this and the increasing length of achievable fully-automated tasks?
On the surface it looks like it's "success at full task automation is much less correlated to success at human augmentation than you might expect", but curious for your take.
English
(Just to clarify, at any given time we way many more than two projects running in parallel, I’m highlighting these two because they were similar lol)
English
METR a few months ago had two projects going in parallel: a project experimenting with AI researcher interviews to track degree of AI R&D acceleration/delegation, and this project.
When the results started coming back from this project, we put the survey-only project on ice.
METR@METR_Evals
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
English
Jeremie Harris retweetledi
Right now, we know a lot about frontier AI development because companies voluntarily share this information. Going forward, I think we need a policy framework that guarantees this. Read more in this oped: nytimes.com/2025/06/05/opi…
English
Jeremie Harris retweetledi
The CCP is using stolen U.S. data & our open system to fuel authoritarian tech—boosting their surveillance & threatening U.S. innovation.
selectcommitteeontheccp.house.gov/media/press-re…
English
Jeremie Harris retweetledi
How can America stay ahead of the AI race?
@GladstoneAI Co-founders @harris_edouard & @jeremiecharris join us with their answer!
English
English
@mkratsios47 @WhiteHouse @POTUS Let us know if you’ve got any feature requests ;)
x.com/AlecStapp/stat…
Alec Stapp@AlecStapp
🚨 NEW TOOL FROM @IFP: The White House asked for input on its AI Action Plan & was flooded with >10,000 responses. To help, we built aiactionplan.org — a searchable database of what America thinks the gov't should do on AI. Naturally, we used AI to make it... 🧵 (1/10)
English
Overwhelming engagement from the American people on the @WhiteHouse AI agenda.
We received over 10,000 comments in response to our request for information on the development of the AI Action Plan called for by @POTUS.
Read them all here⬇️
nitrd.gov/coordination-a…
English
@ChrisPainterYup @METR_Evals @joerogan @harris_edouard Thanks for doing the work, they're some of the few evals that are robust enough to be relevant in that context.
English
Cool and kind of wild to see @METR_Evals's work on agent task length doubling-times mentioned in the opening moments of @joerogan's 3-hour episode on AI with @jeremiecharris and @harris_edouard
English
Jeremie Harris retweetledi
Jeremie Harris retweetledi
Anthropic is preparing for virtual collaborators, the successor to agents. That's what's coming next: security teams and security companies will have new challenges but can also defend with this tech. Had a blast talking with @samsabin923 for this story! axios.com/2025/04/22/ai-…
English
Jeremie Harris retweetledi
U.S. breakthroughs in superintelligence targeted for theft by China's spy network trib.al/Y6CtXu2
English
But as the national security stakes rise, we need to account for CCP state coercion.
In the run-up to AGI/ASI, no picture of U.S. national security is complete unless we explicitly account for the reality of insider threats specifically from the CCP.
English
To be clear: the biggest victims here are the Chinese diaspora targeted by these ops.
Chinese researchers have played a crucial role in advancing American AI.
English
Personnel security for frontier AI is one of the biggest elephants in the room.
According to frontier lab insiders we spoke to, a significant fraction of lab employees have ties to mainland China.
These ties make them easy targets for the CCP state coercion machine.
Billy Perrigo@billyperrigo
🚨Exclusive: All AI datacenters are vulnerable to Chinese espionage, new report says time.com/7279123/ai-dat…
English