undefined | Better HN

0 points_fat_santa2y ago0 comments

By this logic, Priya (the subject of the authors post) would not loose her job but rather just become more productive at her job. And I think that's the key to this whole "GPT4 is taking our jobs argument" that people need to understand, GPT is a bit like Autopilot in a Tesla, it can do the work for you but you still need a trained person there to guide it.

Folks like Priya will not loose their jobs IMO, the company would still need them to provide that last line of validation on data because you just can't blindly trust GPT (I don't know if we will ever be able to bindly trust it), rather the scenario I see is that their team at work would actually be able to stay on top of their assignments instead of constantly being overworked.

0 comments

saulpw2y ago

With GPT-4, Priya can do the work of 10 data curators in the same amount of time! She won't lose her job, they can afford to pay her more, because they won't need to pay 9 of her colleagues. Unless of course, one of her more experienced, connected, upper caste colleagues gets the 10x job instead.

surgical_fire2y ago

Bulldozers also can do the work of 10 dudes with shovels.

Accounting clerks with excel can do the job of 10 accounting clerks with pen and paper.

nr2x2y ago

Hey now, don’t make anybody actually feel responsible for this. We’re just computer nerds

s__s2y ago

Today you need a trained person to guide it. However I’m not convinced that will remain true in the future.

The validation of work is interesting. It does seem essential but it also occurs to me that this would be an absolutely horrible and tedious job. I wonder who would actually be interested in, for example, spending 8 hours a day code reviewing a an AI’s output?

ilaksh2y ago

He knows she will lose her job because he literally said they don't need her anymore since he verified that he can do it with GPT4.

And the new FSD in Tesla only requires people to supervise closely mainly to avoid getting sued. If that system made a lot of severe errors then it would not be deployed.

We have working self-driving in Cruise and Waymo. The only reason to think we really need humans driving at this point is denial. Actually if we switched to AI drivers entirely with the existing systems that would make the roads vastly safer. Accounting for the most erratic behavior of human drivers is one of the biggest challenges.

j / k navigate · click thread line to collapse