All of Lorenzo's Comments + Replies

Does LLaMA have any weird/unspeakable tokens? I've played around with it a bit and I haven't found any (I played with it for a very short time though).

Here are the LLaMA tokens, if anyone's curious. Sadly I couldn't find anything as interesting as " SolidGoldMagikarp"

A lot of this depends on how I approach it: is this something I should work on after the kids go to bed, when I typically write blog posts? Or should I consider trying to go part-time at work, take leave, or quit? I haven't yet talked to people at work about this, but I would lean towards taking leave or going part time: if this is worth doing it's probably worth focusing on. That I think what I'm currently doing is valuable, though, means that there's a higher bar than just "does this seem like a good book to exist."


I wonder if it would make sense t... (read more)

Thank you for sharing this, I was wondering about your perspective on these topics.

I am really curious about the intended counterfactual of this move. My understanding is that the organizations that were using the office raised funds for a new office in a few weeks (from the same funding pool that funds Lightcone), so their work will continue in a similar way.

Is the main goal to have Lightcone focus more on the Rose Garden Inn? What are your plans there, do you have projects in mind for "slowing down AI progress, pivotal acts, intelligence enhancement, etc."? Anything people can help with?

That part was a bit unclear. I guess he could work with redwood/conjecture without necessarily quitting his MIRI position?

5x5 is way too tiny, do you play with a komi of 24 or do you use it mostly for teaching new players?

Edit: it seems both, nevermind

New players, yes; I should have been clearer in my intro. It can also be fun for a bit with experienced players: it took me and my dad a few dozen games before black always won (and lifted all white pieces).

Someone on Twitter mentioned slave owners similarly "not just trading" with slaves who could talk. I think it's a better analogy than factory farmed animals.

Does what we do to factory farmed animals count as "trading" feed and shelter in exchange of meat, eggs and diary?

7the gears to ascension5mo
yeah non-vegans definitely don't just "trade" with cows

I loved the link to the "Resisted Technological Temptations Project", for a bunch of examples of resisted/slowed technologies that are not "eating sand", and have an enormous upside:

  • GMOs, in some countries
  • Nuclear power, in some countries
  • Genetic engineering of humans
  • Geoengineering, many actors
  • Chlorofluorocarbons, many actors, 1985-present
  • Human challenge trials
  • Dietary restrictions, in most (all?) human cultures [restrict much
... (read more)

But weightlifting and working longer hours would still help in both explore and exploit, right?

Why do you get the impression that Adderall and similar don't help with exploring for 50 hours a week? e.g. Erdos explored quite a lot

which I thought was a total waste and if that was what they were going to do they should have donated to LTFF in the first place

I don't understand what was wasted. Do you estimate the costs of running the lottery to be significant?

I don't think there are any extra financial transactions, if you would donate from the same platform in both the lottery and non-lottery cases. (In both cases you donate to EA funds / GWWC, and they then donate it to the org, so 2 transactions)

In theory, yes, but I don't think it has ever happened in practice.

It has happened that you never find out P because the winner doesn't reveal their recommendations, though

Thanks for the reply!

I haven't tried PyPy for this particular project, but my experience previously had been that while I usually got a bit of speedup it wasn't typically much.

That's also my experience in most cases, but in others can be much faster. It does especially well on code with lots of looping that can be JITted.
As a data point, is ~3x faster on my machine (takes 0.8 seconds vs 3 seconds on 100k lines of length 151 containing FF).
Which probably is not enough to make a difference, but might still be useful.

Yes, probably not enough to make a difference; that one in particular is fast enough in python. But useful to have the number!

Probably dumb questions:

 - Have you tried PyPy? It might increase the number of cases where python is good enough.

 - Why C instead of C++? (I assume Rust would slow down prototyping speed because of the borrow checker). Is it because you're more familiar with it from your side projects?

You can also use Numba [] to speed up loops.  It's still slower than C, but it's much better than plain Python code, and it's really easy to implement (just import  numba and put a @numba.njit() before your function).
I haven't tried PyPy for this particular project, but my experience previously had been that while I usually got a bit of speedup it wasn't typically much. I used to work in C++ and know it pretty well, but I don't really like it very much and there isn't anything in C++ I've needed here.

An email was sent yesterday confirming the meetup tomorrow:

I hope it's fine to post it here, otherwise feel free to delete the comment:

On Saturday, we're having the first "official" Amsterdam ACX meetup — exciting! Unfortunately, Scott has not responded to our last emails, so I don't think he'll be there. 

The weather looks good, so we can keep the Westerpark location (in front of Ijscuypje). I will be holding a sign. 

We'll probably be sitting in the grass, so please bring blankets if you have some. 

There's no agenda because I think peo

... (read more)