I do think it deserves to be called quiet. For instance, it seems like they waited until the peak of the news cycle about their conflict with the US government to release this update, and I suspect that was intentional, and also that this worked. In the same week they dropped their core safety commitments, Anthropic was mostly hailed as a hero for standing up to the government; they got almost entirely good press.
But also, Holden’s post explaining the decision is around as understated as a post like that could be. He tried to frame it as something closer to “just another update,” and it was not even the central focus of the post (which I really think it ought to have been, given the gravity of it). The fact that Anthropic was reneging on the core promise of their RSP was systematically downplayed, as it has continued to be by many Anthropic employees who maintain that dropping all “if-thens” from their if-then framework does not meaningfully constitute violating it.
I don’t want this to be a semantics argument about what the word “quiet” means. I will only claim that the Holden post is an important piece of information, both for encouraging Anthropic to be more open and for being evidence against the claim that Anthropic did not want people to know or talk about the relaxation of its safety framework, and further that the description of Anthropic’s behavior as “quiet” gives people a skewed picture of events and does not encourage such prosocial aspects of Anthropic’s behavior.
For instance, it seems like they waited until the peak of the news cycle about their conflict with the US government to release this update, and I suspect that was intentional, and also that this worked.
It happened a few days before the “peak”—it was at a point where barely anyone was paying attention to that particular conflict. Would bet against this being strategically timed[1] at 5:1, if we could reasonably operationalize and figure out a sufficient resolution criteria. (My current model is that Holden is the owner of that project, and would have been responsible for pressing the button on the announcement, and I don’t believe Holden would do that, or knowingly take instruction to do that. Most of my probability mass on what you said being the case lies in worlds where someone else was responsible for the timing of the release, somehow.)
Would not bet against a claim of the form “they didn’t change the timing of the announcement the way they would have done if the announcement had been about something they wanted to see discussion of, like a new model release”.
fwiw I don’t think they “quietly” removed their commitment to pause development, Holden wrote a big LessWrong post justifying the recent changes.
I do think it deserves to be called quiet. For instance, it seems like they waited until the peak of the news cycle about their conflict with the US government to release this update, and I suspect that was intentional, and also that this worked. In the same week they dropped their core safety commitments, Anthropic was mostly hailed as a hero for standing up to the government; they got almost entirely good press.
But also, Holden’s post explaining the decision is around as understated as a post like that could be. He tried to frame it as something closer to “just another update,” and it was not even the central focus of the post (which I really think it ought to have been, given the gravity of it). The fact that Anthropic was reneging on the core promise of their RSP was systematically downplayed, as it has continued to be by many Anthropic employees who maintain that dropping all “if-thens” from their if-then framework does not meaningfully constitute violating it.
I don’t want this to be a semantics argument about what the word “quiet” means. I will only claim that the Holden post is an important piece of information, both for encouraging Anthropic to be more open and for being evidence against the claim that Anthropic did not want people to know or talk about the relaxation of its safety framework, and further that the description of Anthropic’s behavior as “quiet” gives people a skewed picture of events and does not encourage such prosocial aspects of Anthropic’s behavior.
It happened a few days before the “peak”—it was at a point where barely anyone was paying attention to that particular conflict. Would bet against this being strategically timed[1] at 5:1, if we could reasonably operationalize and figure out a sufficient resolution criteria. (My current model is that Holden is the owner of that project, and would have been responsible for pressing the button on the announcement, and I don’t believe Holden would do that, or knowingly take instruction to do that. Most of my probability mass on what you said being the case lies in worlds where someone else was responsible for the timing of the release, somehow.)
Would not bet against a claim of the form “they didn’t change the timing of the announcement the way they would have done if the announcement had been about something they wanted to see discussion of, like a new model release”.
Removed the quietly and linked to Holden’s post, thanks!