I've lived in USA and worked at OpenAI since May 2022, first as resident, then doing research, in reinforcement learning team (which later ended up becoming the whole ChatGPT-research organization), then alignment/superalignment on scalable oversight, now safety systems - model safety. I went to OpenAI because I wanted to contribute to AI safety and that's still what I want to put my work energy towards. I'm on an H-1b and waiting in the green card queue.
The list here is just quickly put together from gpt4o plus light filtering, to generate more options. As of 2024-11-01, my top guess would be OpenAI (by priors even if nothing else), then orgs where I could do relevant AI safety. Unemployed might happen if I happen to be between places when market closes - but given H-1b rules, that would likely not be a long timespan. My immigration situation might also unexpectedly get better. Or I might decide to leave the US.
2024-11-01 is of course in the only correct date format in existence, defined by ISO 8601.
1,000@BobLoblaw hmm good question
If I picked "I'll pick all that applies" that would be giving as much credit to "huge conglomerate" and "tiny 3-person startup they acquired last month".
I'd normally go with "if I work somewhere that got acquired, I'll resolve to the acquiring entity", but I've sorta set myself up by splitting Google deepmind vs other alphabet.
If I went with "will resolve to most specific true option" I'd be setting myself up for "senior meat packing technician on team 4 at the meatpacking plant in Greater Middleton, Wustershire".
Thinking about it, I think the last one is the one I dislike the least. Might get pretty spammy and might be harder for me to aggregate ("come on market will I work in AI safety or not? can you not answer in the form of 300 options with tiny percentages?"). But I hate that idea less than the other options.
Not committing to that yet. If anyone has other ideas how to set the rules in a good way, lmk.
Maybe if we only care about straightforward "X acquired Y", I could just set the resolution rule such that in that case Y wins if it still somewhat meaningfully exists.