From OpenAI's RLHF paper[1]: "By default, when we train a PPO model on our API distribution, it suffers from an “alignment tax”, as its performance on several public NLP datasets decreases." On the HELM[2] site, you can see accuracy benchmarks for InstructGPT <OpenAI model> vs baseline models. The InstructGPT models perform worse on a lot of benchmarks.
It's the "alignment tax". From OpenAI's RLHF paper[1]: "By default, when we train a PPO model on our API distribution, it suffers from an “alignment tax”, as its performance on several public NLP datasets decreases." On the HELM[2] site, you can see accuracy benchmarks for InstructGPT <OpenAI model> vs baseline models. The InstructGPT models perform worse on a lot of benchmarks.
You don't have to care about your job. Most people have about four good hours of work in them. Just do your four hours, and take the rest of the day off. No job is worth burnout.
> You don't have to care about your job. Most people have about four good hours of work in them. Just do your four hours, and take the rest of the day off.
OP says they're young and recently got a high-paying job at a recognizable California company. This isn't good advice.
Working half days every day may work for certain senior people in slow moving parts of a company. However, getting a high-paying job in California at a big-tech company and then putting in half days will get you PIPed and removed really quick.
Or find something to work on that you actually care about. I don't have any trouble working on something for 12 hours a day when I find it intrinsically interesting and meaningful.
Watch that. Doesn't always work. Not everyone you will work with loves what you love or gives as much of a shit about it as you do. And it's painful watching them fuck it up day after day after day. Eventually you end up disliking it by proxy.
I'm also a weirdo. I intensely dislike what I do for a living but I'm good at it and the money is pretty excellent. And I hate it enough to want to do a quality job of it and automate as much of it as possible so I can do something more interesting with my time instead.
> And I hate it enough to want to do a quality job of it and automate as much of it as possible so I can do something more interesting with my time instead.
How does this work in a full time job where you hate what you are doing? When you automate something, you're not freeing up time to do whatever you want. You're just freeing up time to do more of what you don't like, isn't it?
Only if you make the automation public. Plenty of data input guys didn't mention they automated their job and raked in a full FTE working a few hours a week.
Doesn't work as well these days since most knowledge gets shared and a lot of devs are too eager to reveal their cards.
No matter how much you love it, 12 hours a day will never be healthy long term. This is what pro level athletes do to compete and get into the Olympics and they regularly burn out after a few years of it.
I strongly urge you to take the time to establish a long term pace that will allow you to regularly decompress from work, learning, and stress every single day as well as whole “step-away” moments for weeks at a time to allow you to completely disconnect at regular intervals throughout the year.
Based on your description, you’re young and your youth is powering this ability to keep going. Even if you take perfect care of your health, this energy will decline with age, more health complications will arise merely by being older, life will put more expectations on you, or conversely, a lack of life due to grinding too hard professionally will bring depression and loneliness.
Think carefully about how maintainable your perspective on work and life is for what will be at minimum the next 20 years.
> This is what pro level athletes do to compete and get into the Olympics and they regularly burn out after a few years of it.
I believe this is a myth. Excercising for 12 hours a day is not the best strategy to get your body and mind in shape for the Olympics. Internet sources claim that Olympics athletes train 5-7 hours a day. Pro football (soccer) players at top level, with absolutely insane amount of competition constantly breathing down their necks, usually train from morning till lunch and that's it. Doing more would only do harm.
Oh I know it’s a myth, you’re not wrong. I was simply using the myth we keep hearing as a foil for how bad an idea this is. It’s a perfect example of imbalance that results in burn out.
I do appreciate you pointing out that this is a myth. It’s important people don’t think this is somehow manageable, even for “just a few years”.
All that being said… you’re still going to keep hearing it on the news, in books, from movie stars, etc…
I agree that pacing and breaks are important and I didn't mean to imply otherwise. I was offering an alternative perspective to the claim that "most people only have 4 hours of work in them". I just don't think that's true for people who are working on things they find intrinsically motivating. I'm also skeptical of the claim that that 60 hours a week isn't healthy long term, assuming you have the right habits in place. But I'll admit that I haven't really looked into the research on this, if it exists.
Most people don't care about anything that strongly. Among those who do, most of them care about things that others won't pay for (e.g. obscure open source project, or making a hobbit-like garden etc.).
You should find a 100% remote job. If other people are there full time, but you're only there 0-2 days a week, you're never going to be that close with them. You said you're happy when you're at home, but going into the office makes you anxious. So you shouldn't have an issue working at home full time.
David Heinemeier Hansson gave a great talk on air quality and touched upon various HVAC systems. One thing that really stuck out, was his claim that you need 80 CFM per person per room.
https://www.youtube.com/watch?v=MRqh8oLY7Ik
Have you thought about installing a screen door in your bedroom? You could also look into reinforcing it with chicken wire or hail cloth to keep the cat out.
