SB 1047 might have been flawed in any number of ways, but ... it would have made a real difference in the case of xAI / Grok Right?

Would SB 1047 have made a real difference in xAI’s development/deployment of Grok?

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. 🧵

Here's Shortwave | We're Hiring! vs Gemini in Gmail, answering the question of when I last traveled to Brazil (needed for a visa application) Again I say: turn your hyperparameters up! And also ... Google might have doubled tokens / month in the last 2 months, but still has OOMs to go

🧠 How can we solve the unique governance challenge posed by AI? Fathom co-founder and Chief Strategy Officer Andrew Freedman joins Nathan Labenz on the @cogrev_podcast alongside Dr. Gillian Hadfield to unpack how markets might drive better AI regulation. 🎧 Listen now:

Here's me explaining the AI liability legislation I've worked with NY Asm. Alex Bores and RI Sen. RI State Senator Victoria Gu, District 38 to introduce in their states to Nathan Labenz on the The Cognitive Revolution Podcast.

Here's my answer to Nathan Labenz's question on the The Cognitive Revolution Podcast about why I prefer liability to prescriptive regulation as the primary policy tool for mitigating AI risk.

Crypto billionaire James Fickel 🪷's Amaranth Foundation wants is soliciting new ideas on the relationship between Neuroscience and AGI, with emphasis on short-timeline scenarios 👀 Seems extremely well worthwhile for anyone already thinking about this intersection!

Call Launch: In just a few minutes at National Whistleblower Day on Capitol Hill, AIWI and its coalition—represented by Karl Koch in the AI panel—will launch a call with one clear ask: AI companies must publish their whistleblowing policies. PublishYourPolicies.org A thread🔽

The scarcest resource is a concrete positive vision for the future

The race for AI sovereignty is heating up, and the consequences for global power dynamics are enormous. Anjney Midha 🇺🇸, GP at a16z, joins Nathan Labenz on the latest shirley to discuss China's chip independence, MARA, and "a Marshall Plan" for AI.

A lot of people want to work in AI Policy, but entry points are few & far between. Here's a program that pays a competitive wage and is open to citizens of the USA, Australia, Canada, Mexico, and Singapore. Definitely worth checking out!

Fortunately the extra small runs large! AI Digest

NEW: Fresh from the White House, Dean W. Ball joins Nathan Labenz on shirley to break down his role crafting Trump's AI Action Plan—a"deeply positive sum document to come out of a city that is usually quite zero sum" They discuss: * A rare inside look at AI policy-making at the highest

I met Dean W. Ball just before he launched his Substack in Jan '24 – I was immediately impressed with his thinking, and I've also come to trust his character. Dean proves it's NOT TOO LATE for smart, high-integrity people to help shape AI development Get inspired & get involved!

What a podcast from Dean W. Ball -most accurate description of life in centre of govt I’ve heard -how he used LLMs to simulate interagency feedback (genius) -why he “possibly” knew more about what was happening in the labs when outside of govt than in it All that in the 1st hour!

My notes on the podcast with Nathan Labenz and Dean W. Ball, US AI Action Plan author. 👇🇺🇸🤖

@deanbwall is another good get by Nathan Labenz and The Cognitive Revolution Podcast which will be on my playlist this weekend.

I agree 👏