3 ms·
If an AI can do "most economically valuable jobs" retraining won't help. Unless you want to retrain for something that is not economically valuable enough to sp
by oytis 2y ago
If an AI can do "most economically valuable jobs" retraining won't help. Unless you want to retrain for something that is not economically valuable enough to spend compute on it, and has having a human body as a non-negotiable prerequisite.
- physicsguy 2y agoI chose those examples specifically because they’re difficult to replace a physical person there
- disambiguation 2y agoI also aspire(d) to be a physics teacher, but I'm pessimistic about it. Things like khan academy already make "teaching" obsolete, imagine pairing that with an GPT93 agent. 24/7 interactive learning.
- physicsguy 2y agoAll the research I've seen around MOOCs and online only learning is that it's pretty tough for people to stick at, and only 5-10% of people ever complete a course. I just can't see that putting people in front of a computer to do this sort of thing is a good strategy for everyone.
- disambiguation 2y agoOh of course the AI won't be able to baby sit, we'll still need humans for that ;)
- insane_dreamer 2y agoyou're right, it's not but if the cost of doing so is so much less than the in-person experience, then it will become the norm, while the in-person experience becomes relegated to the wealthy like those who can afford private colleges today. And not that many professors will be needed, so good luck!
- oytis 2y agoFrom my experience - that's mostly because courses are useless for career. So attending them is an optional, low priority activity. If you could actually get a respectable degree or skills on a CV that matter for anything other than junior-most positions (if at all), you would see higher completion rate. The quality of material itself is often pretty good. That varies from course to course of course, but I must say I started understanding electronics only after an edx course despite having it in my university.