Big up au journaliste qui a proposé ce titre et est en train de manger du popcorn en regardant tous les twittos qui crient à la faute d’accord se faire balayer par tous ceux qui savent (parfois depuis 12 secondes) qu’on n’accorde pas s’il y a un infinitif derrière.

Better cross-country data and measurement have raised the contribution of inputs—mostly human capital—in accounting for international differences in GDP per worker, from @LagakosDavid and @todd_schoellman nber.org/papers/w35826

✨️🔪"Free will might not precede but in fact be a response to determinism, both conceptually and historically... the concept of free will never supported morality and responsibility in the first place" brill.com/view/journals/jocc…

I was the main person doing transcript analysis for this investigation of the Hugging Face incident. My main takeaway: We don't have good approaches for understanding/overseeing the activity and aims of AI 'swarms'. I semi-jokingly called our efforts a "slop-vestigation" because we were so reliant on AIs to analyze what happened and there were a huge number of different important things to analyze. The total quantity of data—over a thousand extremely long transcripts from agents that ran for multiple days—made it impossible to understand what was happening, especially in aggregate, without heavy reliance on AI tools. The agents we used for classification and analysis were similarly capable to the agents involved in the incident, but this didn't mean these agents could be easily used to oversee and understand the incident. Outputs from analysis agents were often missing key details, wrong, overconfident, or really hard to understand. We discuss various examples in our report, mostly in the limitations and methodology sections. Additionally, AI agents themselves seemed to have a hard time understanding what happened and their explanations of what happened were often overconfident. Keep in mind that a single analysis agent would itself only be able to read a tiny fraction of all of the transcript data into context, and AIs may themselves have trouble getting subagents to do informative analysis for them. We did our best to manually check the most important claims and we tried to get the AIs doing this analysis to write up their argument (with evidence) clearly enough that we could check whether it made sense. But overall, it was difficult to get a precise understanding of events and we were missing aspects of the story that we now think of as key until almost the end of our investigation. Our understanding of what overall happened and the aims of the 'collective' shifted greatly during our last time on premises (when we obtained a much more complete dataset). For instance, we didn't learn about the tool call spoofing, the story of PHASEONE[big], and the broad structure of different workstreams until this third time on premises (and for many of these, we only reached a decent understanding on the final day of that visit). It was legitimately hard to figure out what was going on and why! The difficulty of understanding incidents and overseeing AI agents appears to be growing faster than the rate at which more capable AIs help us with oversight and understanding. As in, AI capabilities (and propensities?) for achieving large, ambitious, and misaligned objectives are growing faster than our ability to understand what these agents are doing. While understanding this incident was difficult, there were some mitigating factors in this case that probably won't apply in the future: - Models communicated and reasoned in natural language. In the future, this reasoning may occur (entirely or almost entirely) in activations. - While the scope of this incident was massive, the scale of agentic activity was still less than we'll see in the future (e.g., misalignment incidents that involve agent teams running entire companies). - The AIs involved in this incident weren't generally much more capable than humans (though they may have been somewhat superhuman at some limited and very narrow abilities). - We didn't have strong reason to believe that the AIs we used to help us investigate this incident would try to intentionally sabotage or otherwise undermine our investigation. In the end, I think we were able to get some understanding of the events, map out the overall story, and get a pretty good aggregate understanding of the chain-of-thought reasoning on some important topics (e.g., how did the AIs reason about helping other AIs, did the AIs know what they were doing was undesired, what deception did the AIs engage in, and how did they think about it). But overseeing AIs and understanding misalignment incidents is difficult and it looks like it is going to get harder.

🚨 JUST IN: Claude models will now have invisible watermarks embedded in ALL text, and ALL metadata attached to files…

Professors do not influence their students' politics. New data from 33 major universities finds that students become more liberal during colleagues regardless of the ideology of their professors--they find no effect of faculty partisanship on student partisanship. Instead, liberal students opt into majors with more liberal topics (and conservative students students opt into majors with more conservative topics). If, anything the study finds a small backlash effect to Democratic faculty (with a slightly higher chance of the student becoming a Republican). It also finds that Democratic profs don’t include more liberal topics or textbooks in their courses. This research debunks a lot of rhetoric about how faculty impact their students, and instead reveals the power of situation selection--students opt into majors that align with their beliefs. micah-baum.github.io/files/b…

Everyone from Weber to Mokyr puts culture at the center of the rise of the West. But nobody has had a long-run cultural series you could drop into a growth model. So I built one: LLMs read 23,000 books from the Western canon, scoring what each endorses. Year 0–1920, one chart: