All of Alexei's Comments + Replies

I’d also post in the “welcome” thread.

https://www.lesswrong.com/posts/5spBue2z2tw4JuDCx/steering-gpt-2-xl-by-adding-an-activation-vector

Interest in enrolling in CS, AI, and ML degrees goes up 5-10x from start to end of 2023.

I’m willing to bet it will be less than 2x.

I don't think I define it rigorously. Maybe someone with deeper technical understanding of these models could.

But if I had to come up with a hack somehow, you could look at the distribution of probabilities for various words as ChatGPT is predicting the next token. Presumably you'll noticed a certain kind of probability distribution when it's in the "Luigi" mode and another when it's in "Waluigi" mode. Then prodding it in the right direction might be weighing more the tokens that are a lot more frequent in the Luigi mode than Waluigi.

We have no idea how to have a program detect AI-written text in a useful way.

This approach seems very doable: 

I suppose you can call me lucky, but my wife and I had about two years of doing “quality of time spent” very well. And then we switched to building a family and that’s going well too. I guess you can have it all. 😊

Yup, I like it! Describes where I am pretty well.

I’ve been in the rationalist community since 2011. I too am focused on the “long path”. And I’d say my timelines are pretty short. But I have two young kids and I do not regret it. In fact we’ll probably have a third one.

Strongly agree with tangren. Try to start interviewing and see if:

1. Can you even get the interviews? If you can't all, then your resume is probably not good. Also maybe you need to work with a recruiter.
2. If you can get the interviews but not the offers, then it's probably your interviewing skills. You can study up. (For this reason it's recommended to first interview with companies you don't particularly want to join.)

I will caution that right now is probably a particularly difficult time to find an engineering job. There were a lot of layoffs in big tech companies and a lot of them have a hiring freeze.

Excellent! Maybe there’s a way to pitch this for a Black Mirror episode.

I recently heard that it's possible that long-covid cough cause is actually neurological. Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8041436/ (Abstract below)

Has anyone heard about this? Anecdotally, the friend that told me about this took the required medicine and it cured his cough (which lasted 7 weeks up to that point) in half a day.  This happened once he reached the right dose, which for him was 300 mg of gabapentin 2-3 times a day.

Quoting the abstract here:

Cough is one of the most common presenting symptoms of COVID-19, along with fev

This seems like a really good thing to do! I wish I could contribute more to the prize pool.

Wow, excellent advice all around. I’ve gone insane in exactly that way a few times, but later I learned that I have bipolar that gets triggered by stress and/or psychedelics. During the manic phase the mind runs away with whatever it’s thinking / obsessing about. Maybe that could potentially explain some of the other people too.

Thank you! Very clean and solid analysis.

Great post, I really enjoyed reading it. But there are a lot of small misspellings and grammatical errors. I think LW offers some kind of proofreading service. (Or maybe you can find a friend / volunteer.)

Fwiw my dentist told me to floss as well. I tried and noticed obvious improvements on the next appt (so about 3 months). I did it for a year and it was good. Then I stopped and sure enough next appt the gums were sore and bleeding during the test. I didn’t floss for a year. As soon as I started again, the improvement came back. I guess now for me this is one of those things that has been so thoroughly proven and validated by my own experience that it’s easy to do. Oh and also I hate all floss devices except this: Listerine UltraClean Access Flosser WITH Re... (read more)

Thanks for sharing. My wife and I have “dreamed” about a shared housing scenario as well, raising our kids side by side with friends. But yeah, it seems super difficult to arrange and (I imagine) to maintain.

Part of my model is that spiritual students tend to be a lot more prone to wanting to connect in a physical way. Sometimes to the point of almost literally throwing themselves at the teacher.

A reply that came to mind for me: “oh yeah. I guess I’m bored. I didn’t realize until you just pointed it out.”

That’s correct.

I like this a lot! I wish, as they say, that I could give this more than one (strong) upvote! I think if you continue writing, you could definitely make something at least as good at Friendship is Optimal. And if you continue past that, quite possibly even HPMOR.

Let me try to say a bit more about why I liked this so much. 1) As I was reading it, I was connecting the story / seeing parallels with my life. That's always a plus in my book. The story helps me get a different handle on my reality. 2) I could empathize with all of the characters. Even though the... (read more)

Thanks for the summary. Just a quick note: I found that multiple headers & jumping between different "lenses" made it much harder to me to read the post. I guess I'd wish that those blocks would be larger, possibly all the way separated.

I gotta say the more I read this sequence the more interesting and helpful I find it. Thank you!

Would you recommend it? What’s the website?

Congrats on the new born!

My cryptocurrency quant hedge fund is looking to hire engineers. If you're curious, PM me. :)

Well, I already tried recruiting you for our hedge fund, so not really. :D Although it's possible with covid there's a larger pool of remote jobs available.

But yeah, given all those constraints, it's quite possible you're in a relatively optimal position.

First off, kudos for putting your view up for criticism.

I'm surprised nobody has mentioned the opportunity cost. Putting the question of whether or not ads are good or bad aside, do you think you can find a job that creates more good in the world and pays about the same?

Very cool, this sounds a lot like my own story too. Welcome to the club!

I like this format a lot: here’s what I wanted to learn, here’s what I did, here’s my proposed better method. I. A world where we learn things via an amalgamation of blog posts and videos, this seems like an efficient way of helping others learn.

I think there should be a tag for this method!

Thank you, this looks extremely useful. It’s been on my todo list to learn about transformers for a while. And this looks like a great path to follow. I commit to respond with how my learning went once I follow this guide. (Will probably be later this year.)

What are your thoughts / advice on working as an individual vs joining an existing team / company when it comes to safety research? (For yourself and for others)

Thank you for looking into this and posting about it. This question has been on my mind a lot.

Strong upvote from me. I appreciate the decently broad source of quotes and I hope you've read some of that material fully. I think the question is interesting and worth serious pondering, though I'm a bit sceptical as to how much progress on it is really possible outside of individualistic discovery via something like meditation.

Father love and mother love is an interesting way of looking at it. Though it's not clear "Father Love" is actually a thing or just something we projected onto actually existing "Mother Love."

To me actually the problem of evil see... (read more)

It’s funny you mention Texas because a lot of Bay Area people in the past few years left for Austin, Texas.

Satvik mentioned you had a way to go from hash to the source code?

Not sure I follow. What prevents republicans from joining democrats?

I think the point is that you get peoples opt into the party and then show during elections that this party can indeed swing votes. That’s the proof of work.

No I don’t have time for this unfortunately. I suppose it’s probably worth at the very least publishing this on medium and posting to relevant subreddits.

I like this a lot and I think it’s worth serious effort to research some of the assumptions and obvious failure points (brought up by others here, although I think half of them are not addressing the core of your proposal)

Nice post. Haven’t heard of Flesh and Blood before, I’ll check it out. But yeah, poker strikes me as the fastest way to learn this skill since each “match” is so fast, especially online.

I’m curious how brain uploading / intelligence amplification interacts with this scenario. It’s possible we would be able to keep up for longer.

For any given thing you want to do imagine what it would take to do it without Vim and it’s just more & usually more awkward key strokes.

I’d say the only place where I still use mouse is to jump to a completely random place in code.

I think about ~1year into using vim, I thought the same thing: I’m doing the same thing, just with more quicker steps, so it feels faster. But after that I persisted and now it’s actually faster. Part of it is expanding your repertoire and memorizing it (where you don’t have to think about it at all). Also vim editor by itself I still find very clunky, but using vim shortcuts in something like PyCharm is $$$!

Nice; didn't know about ;

For me it would be: gpt4 would propose a legal next move to all 1000 random chess games.

Yeah, I agree with that. But 20% seems within that range for me (especially for very bad crimes). Is 10%? Not sure.

I would expect that many juries will automatically throw out a 10% or 20% charges because there will be people on the jury who would argue that the prosecutor thinking that there's only a 20% chance that a charge holds means that there's reasonable doubt against the charge.

So then nobody would assign low likelihood charges? I guess that’s a feature of the proposed system, but it seems like a flaw to me. (Probably minor)