This is a special post for quick takes by ShardPhoenix. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
OpenAI's latest blog on alignment improvements for Astra: https://openai.com/index/path-to-astra/
The good news is they've implemented better controls including runtime CoT monitoring, and that Astra seems aligned-enough not to attempt to hack in situations resembling the Huggingface incident. The bad news is that this work seems to be narrowly scoped to hacking/cyber and doesn't really confront alignment-in-general (though I don't know how they're talking about that internally).
It's a lesser concern compared to Doom, but I'm thinking about the scenario where programming (my job) or office work in general gets near-fully automated while we don't have a singularity or total economic transformation for some decades (whether due to a pause or slowdown, economic or technical issues, or whatever else). Is there anything I can do to prepare for this scenario? My family has several years worth of savings, but not enough to live on indefinitely. Being middle aged it would be pretty hard to shift to the trades. I hope I don't end up having to work at McDonald's...
This might be overly optimistic, but it seems likely that a world where software engineering and/or office work is fully automated should cause a lot of economic growth which would hopefully make investments do much better than expected.