Geneva KD

AI Policy @JoinFAI

[Zach Graves 🍞🎪 @ZacharyGraves]

Excited to share this updated primer on the Office of Technology Assessment (OTA) with @danielschuman. With new economic/cyber/bio/and other policy challenges arising from AI, now would be a great time to invest in greater science and tech capacity in Congress.

cc @SpeakerJohnson @hakeemjeffries

[Chris Harig @chrisharig99]

The US-China SI hotline is a great first step towards mitigating the national security risks presented by SI. @GovindPimpale and I argue that a few design choices would make the hotline a durable institution, unlike it's predecessors. Post linked below.

[Geneva KD @genevese_]

@chrisharig99 @GovindPimpale https://t.co/UNdU0tZ2mu

[Geneva KD @genevese_]

@chrisharig99 and @GovindPimpale argue that the new US-China Super Intelligence hotline is a major step forward, and outline how clear risk frameworks, technical expertise, and durable communication channels can ensure its success.

Policy Gradients piece linked below.

[Geneva KD @genevese_]

@RothRottweiler @JoinFAI @jachiam0 << Tactical spelling mistakes to signal human-written tweets

[Geneva KD @genevese_]

I met with @JoinFAI's newest Non-Resident Senior Fellow @jachiam0 in Berkley last week to discuss the idea of an Acceleration Emergency, mapping the frontier of scientific discovery, and federal jobs guarantees.

You can read the highlights of our conversation below.

[Samuel Hammond 🦉 @hamandcheese]

Just over a week left to apply for the best AI / SI policy fellowship in DC ➡️➡️

[Samuel Hammond 🦉 @hamandcheese]

New on Policy Gradients:

5 questions with @JoinFAI senior fellow @jachiam0!

[Joshua Achiam @jachiam0]

I really enjoyed this conversation with Geneva! Many interesting things to think about for what the next couple of decades will contain.

[Geneva KD @genevese_]

@StephenLCasper @KamacheeMax @KamacheeMax yes 👏👏

[Geneva KD @genevese_]

https://t.co/HRM9MqiRXI

[Geneva KD @genevese_]

@jane_gatsby @ElizabethHolmes Acquit the 4 A’s!

[Abdulrahman Alfozan | عبدالرحمن الفوزان @alfozan]

@SAlsaif7 @bmmashat "The weakest link in the security chain is the human element" 🙂

[Soham Mehta @SohamThoughts]

The best place to put data centers might be the middle of the ocean.

After studying and corresponding with @_panthalassa, a startup building wave-powered data centers, I’m releasing a model and interactive dashboard that tests the viability and cost of moving compute to sea.

If you’re curious about what harvesting massive amounts of energy from the ocean looks like and why I believe ocean compute beats data centers on land, read it.

[Max Bodach @maxbodach]

You asked, we delivered. The @JoinFAI Field Guide to the Tribes of Silicon Valley is live now.

27 tribes, a few easter eggs, and a 3-minute quiz (just for fun).

https://t.co/btQQ9zYJzK

[Farrell Gregory @efarrellgregory]

SINOMODERNITY

assuming symmetry between chinese/american perspectives on machine intelligence is already insufficient

expect chinese views on the use and potential of machine intelligence to become more distinct and as such worth rigorously understanding

read @palladiummag:

[Foundation for American Innovation 🇺🇸🚀 @JoinFAI]

The American Innovation Gala kicks off tonight! See you there 🚀

[Ben Hayum @BenHayum]

Confirmation of what was expected: GLM 5.3's safeguards as an open weight model are extremely weak.

Interesting bits:
- Abliteration (removing safeguards) cost only $4,400.
- Abliteration took GLM 5.3's refusal rate from above 90% to about 3% and 2% on two benchmarks and to 12% on the third.
- Abliteration barely reduced model capabilities. Same scores on GPQA and only 4% reduction in cybergym.
- Even with no abliteration, jailbreak methods were still very effective.
- None of the techniques to bypass safeguards that worked on GLM 5.3 worked on Claude models.

https://t.co/3Ic36V3aAE

[Geneva KD @genevese_]

@PabloAntonio @arenamag @aquavoice Good thoughts accompanied by my favourite Rockwell