This is a special post for quick takes by Sherrinford. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
My understanding of some things in AI development, which is probably completely wrong:
OpenAI wants to automate AI research.
This is bad for AI researchers, including those working at OpenAI, unless they own shares in the company.
The decision-makers have wriggled out of the non-profit status.
This could potentially mean that they are inclined and have the ability to evade other obligations as well.
If I were an AI developer at OpenAI without a deeper understanding of company structures, corporate law, etc., I would wonder if the decision-makers there could find a way to ensure that they ultimately pocket all the money and I get nothing, even if I currently own shares in the company.
given all the known unknowns and unknown unknowns around this stuff, plausible enough to me that anthropic RSI techs are in the same bind.
Like yes based on public information the story that openai RSI techs are in this bind is much cleaner, but I expect like 80-90% of the relevant knowledge to be private.
GDM, idk-- they seem less RSI pilled than the other two (but again, what do I know)