Lennart Finke

[Lennart Finke @LennartFinke]

Read it here → https://t.co/ukAXW7G4bf

[Lennart Finke @LennartFinke]

New short paper: If we make an international treaty to limit AI development, when should it end? We want the benefits of AI eventually, but need time to prepare the world for it. It's an unsolved problem how the treaty ends at the right time by design.

[Lennart Finke @LennartFinke]

Thanks to @aaronscher, @DavidSKrueger, @MauricBaker, @OliverEGuest, @matthijsMmaas, @thlarsen, @So8res for comments! Mistakes and bad takes are mine.

[Cas (Stephen Casper) @StephenLCasper]

🚨 New paper: Some, but not all, AI companies make corporately-loyal models.

xAI, DeepSeek, Anthropic, & OpenAI models all downplay company controversies. Google, Meta, & Alibaba models don't.

The findings are clear, but we are pretty confused as to why... 🧵

@LennartFinke

[Lennart Finke @LennartFinke]

→ https://t.co/NWJRQvZ0PI

[Lennart Finke @LennartFinke]

It turns out some arXiv subject categories have much longer wait times (from submission to announcement) than others, see the blog post ↓ for which are which.

[Cas (Stephen Casper) @StephenLCasper]

@LennartFinke and I will release a paper on Monday about how some AI developers tend to make models that differentially downplay company controversies.

Below (🧵) is a link to a 1-question Google form for you to guess the results before they're out. (They might surprise you.)

[Lennart Finke @LennartFinke]

New piece in @asteriskmgzn → https://t.co/XY41pqoNJM