rramach

I think we all need to update our priors on what LLMs can do today. Here is one anecdote:

Scott Aaronson says "Dana tells me she now mostly understands the proof of the UGC, and is amazed by the new ideas in it, and wants to give talks about it soon."

Dana Moshkovitz is a professor at UT Austin who is an expert in the area and has been working on solving this exact problem for decades!

True but the explanations will get better.

Meanwhile you can use the model to help you out as Scott comments "Just now, however, Dana tells me that she’s been asking Astra all day to explain the new proof of the UGC to her and it’s been doing an amazing job and she’s starting to understand the construction."

No one knows the answer. However, trusting your curiosity and going where it takes you may be a reasonable strategy. If the model asymptotes, you will be ready to figure out how to build upon it and if not, at least you will have had fun satiating your curiousity!

Terence argues that explanation of results ("understanding") will be the new bottleneck in math research but I am not sure this is the real bottleneck for progress.

Understanding was critical for the field to progress when only humans were involved but if humans are not needed to make progress, I wonder if we split into two worlds: an AI math-world where amazing new results continue at a rapid pace bottlenecked only by compute/cost and a human math-world where we understand a subset of the AI math-world as a hobby (similar to Stockfish vs human chess).

Wow, this comment thread clearly shows that at least Anthropic has not been a great communicator.

If one reads this with a charitable lens, Dario is simply saying that 1) Nation state actors are a threat which needs to be combatted by chip bans and distillation prevention and 2) open-weight models can pose biological risk.

One may or may not agree with item 1 but item 2 above should have broad support given the unknown unknowns in play?

Greenwald was not claiming he is breaking the story. He was trying to shed light into the media blacklisting of the story. It is indeed Ironic that his story got blacklisted as well!

I think some editors are perhaps trying to avoid a repeat of the 'Hillary email/Comey announcement of 2016' scenario but didn't anticipate the Streisand effect.

In hindsight, best approach may have been to cover the story, get a sound bite from Biden denying that he benefitted, say there is no evidence he benefitted and leave it at that.

I don't get this obsession with comparing Sweden, a remote, sparsely-populated country, to more dense, well-connected, large population centers. If you really want to compare Sweden, please restrict your comparisons to its closely similar, equally remote and sparsely populated neighbors, namely, Norway and Finland. The 10X difference between them will be immediately obvious.

Yes, I agree. This issue is super-nuanced and I believe Mark has spent years (since the 2016 election fiasco) looking into this as well as being advised by top minds to arrive at his current position. These employees are acting emotionally now but will realize the value of Mark's position over time (in my opinion).

John's analysis is cherry picking in many ways. Death rate in Diamond Princess is 1.1% today and 2% is listed as severe. Assuming 50% of severe make it, final fatality rate for the ship may end up closer to 2%. John then adds a 50% discount factor but it is not clear how he picked that number. Also, the 1% of population infected seems to be another number pulled out of a hat. If we are basing our figures based only on the ship with no other assumptions, we have to go with 20% infection rate. Thus, one reasonable estimate of risk from the ship data is 20%x2%x330M = 1.3M deaths if we wait for "evidence" and did nothing. Clearly, this argues for doing something!

Edit: Also, Germany does not test dead folks for coronavirus while Italy does. Further, SK death rate has gone up to 1.3% (0.9% is an old number) and many more are in severe category. Thus, the sub 1% numbers seem more like the outliers than the above 1% numbers.

Classics have a way to taking things to extreme to illustrate various points. The Gulag example highlights that when a person had a choice to take the road to the left (selfish) or the right (selfless), they took the right road and gave up their self/happiness but ended up achieving meaning. Frankl's book on search for meaning also talks about meaning being more desirable than happiness.

The point is not that suffering makes ones life meaningful but the struggle or striving or the road seems to be preferred by the wise than an end state like happiness.