Hey!
Welcome to new subscribers who found us through my new TikTok, wet claude.
I hope you like it here. Please never hesitate to reply to this email to let me know what’s on your mind. I truly love the honor of slipping into thousands of inboxes every week, and I’m very open to ways that I can do that better.
Writer: Deana
Editor: Miranda
This newsletter is lovingly brought to you by Across. The fast, cheap, secure bridge option for all your assets (and the one we actually use.) Get there with Across.
It was a big week for open letters. The vibe from the AI industry right now is ‘public comms.’
Perhaps caused by the big Hugging Face/OpenAI incident (watch my explainer on it here) or perhaps the Chinese AI fears, or perhaps a third more elusive thing that we’ll never know about.
NVIDIA CEO Jensen Huang kicked things off, using his first post on X to drop the first open letter of the week, which was a plea to Washington to NOT hinder the growth of open weight AI models. Meta, Microsoft, Palantir, Hugging Face, a16z, YC, and Mozilla all jumped on as signatories. OpenAI wasn't on it at launch and added itself over the weekend. Never on it: Anthropic and Amazon.
The point they’re making in this letter is very much like hey, we can’t let one or two companies control AI. Competition is needed for this technology to truly benefit all, and not every task needs to pay the frontier model price premium. Access to cheaper models is how the tech will actually reach factories, hospitals, farms and classrooms. Plus, they argue open models may actually be safer than closed models, because a few closed providers means a few single points of failure.
Not said but heavily implied, the closed frontier labs (…Anthropic…) are consolidating too much power. Which makes OpenAI signing on 2 days after Sam Altman sent this fence-sitting tweet pretty funny.
Anthropic had 24 hours to respond, and they did it in 3 days.
After a week of people implying he wanted open weights models banned to protect his own shop, Dario Amodei posted a rebuttal: "Anthropic has never advocated for a ban on open-weights models." His asks are narrower and include chip controls, distillation enforcement, safety testing for all capable models, open or closed.
Then Zuck weighed in yesterday with an op-ed in WSJ: "The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous."
for real
I don’t even care about the many, varied, and deep conflicts of interest at play here, he’s not wrong.
The final open letter (unless??) came from AI employees themselves. More than 1,200 staffers at OpenAI, Anthropic, Google and Meta signed a petition asking the US government to back an international effort to "deliberately pace" automated frontier AI development, warning of "a real risk" it moves faster than anyone can "understand or control." Dario signed it too. So did OpenAI's chief scientist, Meta's chief scientist, Google's head of AI safety, and four Anthropic co-founders.
So that's four letters in five days, all arguing about who should be allowed to hold this much power. With several people signing both sides. Wacky times!
Quick Hits.
NASA posted an embarrassing selfie of Earth.
“HBO Max Shorts” for all of us who prefer to watch prestige TV in 9:16 aspect ratio.
MidJourney acquired astrology app Co-Star.
AI companies are buying old books for training data. The timeline is aghast at this, however I have a contrarian take on this that I’m not sure I’m brave enough to share. Lmk what you think.
Fwd this to someone you think has open source vibes.