> A 24 year-old AI researcher will earn 327x what Oppenheimer made while developing the atomic bomb.
Where is the contradiction??? Oppenheimer was in no way connected to the "value" generated by his invention. The AI researcher can be part of a small company and have a tiny chance of huge success, so direct connection to the value generated by his work.
Also, it's not like Oppenheimer could get offers in the open market to get a bigger salary or extort other people not to drop a bomb without maybe dying in the process. But if he could, he could hire hundreds of those AI researches as assistants...
Those two examples are government-run projects and in the case of the Manhattan project were part of a scheme with the main purpose of murdering people at a mass scale. The question there was more how much would we have had to pay those people not to do what they did.
These aren't good comparisons for someone who is doing work we expect, in advance, to be a net good. It isn't a particularly powerful comparison - we already might expect that private markets pay better just because people are deployed to useful work. It is actually a pretty reasonable suspicion that this bloke is going to do more than 300x as much good as Oppenheimer, both morally and commercially. Any deaths as a result of his direct work will be accidental.
In some ways it's a stupid comparison because back then people didn't take money that seriously because the state still knew how to hire serious talent and reward them with status instead.
We let that article take precedence on the front page, as the top commenter on this thread makes a good point that the comparisons to the Manhattan Project and Space Race can be regarded a red herring and diminish the weight of the article. The NYT story on the front page focuses on the central topic of comp offers to A.I. talent.
Think about all the useful and necessary things that professional athlete's salaries dwarf.
I feel that at least these researchers are getting paid their perceived value instead of all that value being absorbed by FANNG.
I think AI research is sort of like cutting a diamond, where every few percent more efficient could result in huge sums saved in infrastructure costs for training and inference and Capex.
8 comments
[ 0.19 ms ] story [ 22.6 ms ] threadWhere is the contradiction??? Oppenheimer was in no way connected to the "value" generated by his invention. The AI researcher can be part of a small company and have a tiny chance of huge success, so direct connection to the value generated by his work. Also, it's not like Oppenheimer could get offers in the open market to get a bigger salary or extort other people not to drop a bomb without maybe dying in the process. But if he could, he could hire hundreds of those AI researches as assistants...
These aren't good comparisons for someone who is doing work we expect, in advance, to be a net good. It isn't a particularly powerful comparison - we already might expect that private markets pay better just because people are deployed to useful work. It is actually a pretty reasonable suspicion that this bloke is going to do more than 300x as much good as Oppenheimer, both morally and commercially. Any deaths as a result of his direct work will be accidental.
A.I. Researchers Are Negotiating $250 Million Pay Packages - https://news.ycombinator.com/item?id=44766758
We let that article take precedence on the front page, as the top commenter on this thread makes a good point that the comparisons to the Manhattan Project and Space Race can be regarded a red herring and diminish the weight of the article. The NYT story on the front page focuses on the central topic of comp offers to A.I. talent.
Think about all the useful and necessary things that professional athlete's salaries dwarf.
I feel that at least these researchers are getting paid their perceived value instead of all that value being absorbed by FANNG.
I think AI research is sort of like cutting a diamond, where every few percent more efficient could result in huge sums saved in infrastructure costs for training and inference and Capex.