This is a special post for quick takes by gluesniffer1984. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
I'm a high school student who is currently planning on studying econ next year at a mid tier university, but after lurking here/EA forum for a while I've become very interested in mechinterp/alignment/all of the other wacky machine learning fields.
I am interested in these fields for the obvious reasons (very interesting, may help in reducing extinction risk) but I'm wondering if I even have a shot at being employed if I do choose to pivot from economics to AI safety.
Concerns:
- I consider myself pretty smart, but I'm definitely not smart enough to compete with other newcomers to the field that are the same age as me and have devoted their entire lives to the practice, I.E I'm worried about getting out competed for employment positions by smarter people. Is there still room in the field for people in the, say, 90th percentile of AI knowledge, or just the 98th percentile?
- What is the probability that all work in this field gets automated away anyways and I'm left holding the bag?
Dredged from the cesspool of Linkedin:
It seems like this person was so turned off by the hype around productivity gains that it has now negated the value of recent safety related evidence (hugging face breach, MIRI reports, what have you). Is there a name for this bias?
Seems related to seeing things as black and white, or on a scale, as opposed to nuanced. A halo effect.
The AI can simultaneously make the GDP skyrocket and exterminate all humans. There is no contradiction between that. But for some people, intuitively, the AI must be on a linear scale between "extremely good" and "extremely bad", and getting pieces of evidence for both puts it in the middle of the scale, i.e. "normal, boring".