AI Agenda: Will Reasoning Models Think Clearly for Much Longer?
'Chains of thought,' a sneak peek into the minds of AI models, are slowly becoming gibberish.
Read more from @rocketalignment π
theinformation.com/articles/will-β¦
Ok you know whatβs crazy, I went back to the problem and the answer involved a larger and a smaller carbon ring. And Frederick is apparently in the outer ring of DC suburbs
Thinking Machines Lab's round is unusual in both its sizeβ$2B @ $10B price led by a16zβand its voting rights:
- Mira Murati's vote = # of other board directors + 1, giving her a majority
- Founding shareholders votes have 100x the weight
w/ @nmasc_:
theinformation.com/articles/thinkβ¦
First Gary Marcus column I've truly disliked
[Things companies say] and [things that are true] are obviously correlated, but even on models where they're not, the hype objection is such a red herring
And it's poor form to not even read the article you're critiquing
This is a sign of
1. Maturation for AI: growth has slowed enough and managers have enough experience to start building out a normal internship program
2. Derisking the talent pipeline from dependence on MATS
Weβre starting a Fellows program to help engineers and researchers transition into doing frontier AI safety research full-time.
Beginning in March 2025, we'll provide funding, compute, and research mentorship to 10β15 Fellows with strong coding and technical backgrounds.
Also featuring a classic tale of velvet handcuffs
Three ex-OpenAI employees say they were interested in joining US AISI but couldn't because of their OAI equity. Two of them ended up at Anthropic instead. The equity issue was counterfactual for one of them
AI Agenda: DeepSeekβs Chatbot Outpaces Google, But Google Closes the AI Efficiency Gap
Google is closing in on DeepSeek when it comes to balancing quality and price, while DeepSeekβs chatbot has leapfrogged Googleβs Gemini.
Read more from @amir and @rocketalignment π