LessWrong 2.0 Reader

Sorry if I'm missing something stupid, but doesn't that first sentence there explain the situation? “Please note that we are not accepting applications to use the cluster at this time.” I would presume the Submit button is just vestigial due to not being able to easily hide/remove it at that target URL.

Assuming I'm right, what would be nicer presentation-wise is if the page at https://safe.ai/work/compute-cluster were to change its button to a non-button reading “Applications Currently Closed” or such, and if the explanation included something more explicitly referring to the non-functional UI like “This form thus cannot currently be submitted.”

I really wonder what effects text like this will have on future chain-of-thoughts.

If fine-tuning on text calling out LLM COT deception reduces COT deception, that's one of those ambiguous-events-I-would-instead-like-to-be-fire-alarm-things I hate. It could be trying to be more honest and correct a bad behavior, or just learning to hide its COT more.

I think you would be able to tell which one with the right interpretability tools. We really should freak out if it's the latter, but I suspect we won't.

I guess the actual fire alarm would be direct references to writings like this post and how to get around it in the COT. It might actually spook a lot of people if the COT suddenly changes to ROT13 or Chinese or whatever at that point.

has anyone seen experiments with self-improving agents powered by lots of LLM calls?

Are you genuinely unfamiliar with what is happening to the uyghurs, or is this a rhetorical question?

Gott started this type of confusion than he claimed that Berlin wall will stay 14 more years and it actually did exactly that. A better claim would be "first tens of hours with some given credence"

This sounds like both an alignment and a capabilities problem.

AI 2027-style takeoffs do not look plausible when you can't extract reliable work from models. 

are you saying something like: you can't actually do more of everything except one thing, because you'll never do everything. so there's a lot of variance that comes from exploration that multiplies with your $O\left(k^2\right)$ variance from having a suboptimal zero point. so in practice your $k$ needs to be very close to optimal. so my thing is true but not useful in practice.

i feel people do empirically shift $k$ quite a lot throughout life and it does seem to change how effectively they learn. if you're mildly depressed your $k$ is slightly too low and you learn a little bit slower. if you're mildly manic your $k$ is too high and you also learn a little bit slower. therapy, medications, and meditations shift $k$ mildly.

“In a few rigged demos, it even lies in more serious ways, like hiding evidence that it failed on a task, in order to get better ratings.”

Yeah tbh these misalignments are more blatant/visible and worse than I expected for 2025. I think they'll be hastily-patched one way or another by this time next year probably.

InverseGaussian[7.97413, 1.315]

Can you elaborate on where these numbers coming from? Eli's screenshot for the inversegaussian had parameters of 5.3743 and 18.9427.

Regarding the Russians and East Slavs more broadly, Anatoly Karlin has some napkin math that at the very least shows the huge toll that the world wars had on their populations, which barely grow or s:

(8a) Russia just within its current borders, assuming otherwise analogous fertility and migration trends, would have had 261.8 million people by 2017 without the triple demographic disasters of Bolshevism, WW2, and the 1990s – that’s double its actual population of 146 million.

Source: Демографические итоги послереволюционного столетия & Демографические катастрофы ХХ века by Anatoly Vishnevsky

(8b) According to my very rough calculations, based on various sources, the population change for each of the following in their current borders between 1913/14 and 1945/46 was about as follows:

• Russia – 91M/97M
• Ukraine – 35M/34M
• Belarus – 7.5M/7.7M

Assuming a threefold expansion in all of these populations, we could have been looking to a Russian Empire or Republic with a further ~120M fully Russified Belorussians and largely Russified Ukrainians, for a total Slavic population of almost 400M.

That’s twice bigger than the number of White Americans today, the most populous single European ethnicity, and almost as much as all of today’s Western Europe.

(8c) Total population of a hypothetical Russian Empire that also retained Central Asia and the Caucasus, and that hadn’t been bled white by commies, Nazis, and Westernizers during the course of the 20th century, would likely have been not that far off from Dmitry Mendeleev’s 1906 projection of 594 million for 2000.