Do you know if anyone has done an analysis of how many people are actually working on these risks (especially broken up into different types - ie misalignment, misuse, etc)? I know 80k used to say 200 people but that was like 2021.
80k commissioned some more recent experts from a superforecaster, but unfortunately they haven't been published. I hope there will some better public work coming up.
For misalignment, and from knowing the orgs in the field, I'm pretty confident it's thousands rather than tens of thousands. Many of the other key risks listed have less attention.
I suppose anyone in the education field, particularly teaching, knows that AI is already taking over in so many ways, most of them negative. I teach in an international school and it is a rare student who is not using chatgpt or something similar on ALL of their papers. Even when I ask students to compose an opinion essay in their native language and then translate it, their preference is to just hit a few keys and get an AI generated essay based on their topic. Non-native speakers that I work with tend to view AI as the great equalizer. And it is being sold as such. One need only watch one Grammarly ad to recognize that they are trying to convince folks that doing one's own work and even thinking are things of the past. I know a student who is now at UC Davis who did not compose a single piece of work in any language while she was at the high school from which she graduated. I was a bit shocked when I noted that she was not being asked for any references on her UC application. I don't even want to get into the overseas SAT exams. (Not entirely due to AI but the work "smarter" mentality is definitely encouraged by one's ability to work "smarter" by using AI to do their work.) I would hope that universities are becoming vigilant to the threat of AI completely dominating academics and intellectual thought. It is becoming more and more difficult to spot academic dishonesty and I think it is accurate to say that most private schools tend to turn a blind eye to the use of AI as their bottom line is profit and graduating students into top universities is what attracts folks to pay hundreds of thousands of dollars that they pay over the course of a student's "career" in their institutions.
Really appreciate your work Benjamin. Wondering how you would respond to people like Dwarkesh that say that 80,000 Hours is fond of making career-altering recommendations that change quite often?
I'm not sure I've heard that critique by Dwarkesh. Which opportunities are best will change over time as the situation changes and we learn more. I actually think 80,000 Hours wasn't fast enough to recommend AI more strongly after the situation changed with chatGPT.
Do you know if anyone has done an analysis of how many people are actually working on these risks (especially broken up into different types - ie misalignment, misuse, etc)? I know 80k used to say 200 people but that was like 2021.
80k commissioned some more recent experts from a superforecaster, but unfortunately they haven't been published. I hope there will some better public work coming up.
For misalignment, and from knowing the orgs in the field, I'm pretty confident it's thousands rather than tens of thousands. Many of the other key risks listed have less attention.
I suppose anyone in the education field, particularly teaching, knows that AI is already taking over in so many ways, most of them negative. I teach in an international school and it is a rare student who is not using chatgpt or something similar on ALL of their papers. Even when I ask students to compose an opinion essay in their native language and then translate it, their preference is to just hit a few keys and get an AI generated essay based on their topic. Non-native speakers that I work with tend to view AI as the great equalizer. And it is being sold as such. One need only watch one Grammarly ad to recognize that they are trying to convince folks that doing one's own work and even thinking are things of the past. I know a student who is now at UC Davis who did not compose a single piece of work in any language while she was at the high school from which she graduated. I was a bit shocked when I noted that she was not being asked for any references on her UC application. I don't even want to get into the overseas SAT exams. (Not entirely due to AI but the work "smarter" mentality is definitely encouraged by one's ability to work "smarter" by using AI to do their work.) I would hope that universities are becoming vigilant to the threat of AI completely dominating academics and intellectual thought. It is becoming more and more difficult to spot academic dishonesty and I think it is accurate to say that most private schools tend to turn a blind eye to the use of AI as their bottom line is profit and graduating students into top universities is what attracts folks to pay hundreds of thousands of dollars that they pay over the course of a student's "career" in their institutions.
Really appreciate your work Benjamin. Wondering how you would respond to people like Dwarkesh that say that 80,000 Hours is fond of making career-altering recommendations that change quite often?
I'm not sure I've heard that critique by Dwarkesh. Which opportunities are best will change over time as the situation changes and we learn more. I actually think 80,000 Hours wasn't fast enough to recommend AI more strongly after the situation changed with chatGPT.
Interesting! And it came from a quote during his recent AMA episode.
I don't remember him making that exact critique. Also see: https://x.com/dwarkesh_sp/status/1905130851723276481