Was great to talk with @AU_SIS about how my experience was shaped by my education with AU, and looking back at my service and how I hope I’m still giving back.
Air Force veteran @JasonBakerJB, SIS/MA ’19, spent 20 years carrying out U.S. foreign policy. At SIS, he brought a practitioner’s eye to the classroom. Now he’s working on one of today’s fastest-moving national security challenges: AI. https://t.co/Ql6PNJKGmz
I’m working on a new, bipartisan AI bill called the Safe and Accountable Federal Evaluations of AI Act — aka, the SAFE AI Act. No industry in America gets to police itself, and AI shouldn’t be the exception. We need real, mandatory guardrails now.
https://t.co/qKcl3ecKG2
The Commerce Department approved the sale of almost 580,000 Nvidia AI chips to China under a Trump administration plan to allow the sales for a cut of the revenue.
The customers included companies that the Pentagon lists as having ties to China’s military. https://t.co/FMpXYmVqkX
METR shared the latest version of our "Time Horizons" work today. Takeaways, a few of which were already becoming clear imo, and are just further substantiated by this work:
1. Frontier AI agent time horizons are now doubling roughly every 3.6 months. Naively extrapolating the 50% success rate in this work, we would get months-long AI agent time horizons on well-specified tasks in the summer of 2027. Seems faster than the expectation I described below in February.
2. It's getting much harder to rule out serious AI loss of control risks using capabilities-based arguments (e.g. arguments of the type "I'm not worried that my AI agent will take unintended actions by itself, because it can't do anything by itself!"). Again, was becoming clear based on many sources of evidence, but again reiterated here.
3. The tasks we have to make are now so long and require so much human skill that its difficult to get human baselines for them (as expected).
4. Holding task-length constant, agents do worse on tasks that would have been costlier to make.
5. From @rowanwhuang: "Larger swarms tend to have bigger gaps on their performance on the test set they generate compared to what the hidden tests are actually testing for. This might suggest an “echo-chamber” effect, where swarms convince themselves to over-optimize on bad metrics, and don’t check their assumptions well."
6. When the team tried to have AI agents create their own benchmark tasks that could themselves be included in our time horizon task suite, the resulting tasks were borderline above our bar for a work-test, but not good enough to actually use.
Reuters: The US Securities and Exchange Commission is probing Nvidia's accounting practices, said two people briefed on the matter, adding to the legal woes facing the scandal-hit chip giant.
Went on @BBCNews to discuss with @SumiSomaskanda the recent disclosure by the House Foreign Affairs Committee that over 30% of the Nvidia and AMD AI chips that the Commerce Department approved for export to China went to Department of War-designated Chinese Military Companies.
Wow - Senator Hickenlooper (CO) with a new letter calling on Senate leadership to make AI safety a top priority for the coming Congress, and to consider legislation "banning RSI" to stop "AI systems from autonomously advancing beyond our ability to control them."
Representatives Beyer, Moran, Whitesides and 20+ other members launching a new bipartisan AI Safety and Innovation Caucus.
All three are leaders on this issue and its great to see them emphasize that action on AI safety is too important for normal partisanship to block.
We have learned that third parties who provide services to the firm were the target of a cyber attack.
Based on our investigation so far, we believe the attackers accessed certain marketing data held by those vendors, which may include some customer names and email addresses.
There is no disruption to our services.
Where appropriate, we will contact affected customers directly and offer support, including complimentary credit monitoring.
There is no connection to service disruption some customers experienced earlier this week.
The U.S. can ban China from buying AI chips, but what if they can just rent them online instead?
@michellesnie breaks down the Commerce Department’s new rule that aims to close this loophole. Read her full CNAS Insights piece here: https://t.co/uK6RHff70V
TSMC is tightening restrictions on China following the new US Department of Commerce rules that took effect today. Orders using manufacturing processes of 16 nanometers and below have been halted unless customers obtain US export licenses. The Chip War is slowly ramping up.
EXCLUSIVE: A huge cyberattack has hit the European Commission.
The incident has potentially compromised swaths of personal data and prompted the most senior officials to discuss the crisis at their weekly meeting.
More in Brussels Playbook 👇
https://t.co/9WURHuiJLZ
Pottinger is absolutely correct. The idea that Beijing would ever have simply sat by and allowed semiconductors to be a chokepoint on their technological development is a complete fiction, and directly contradicted by explicit statements of national policy by Xi Jinping himself.