Apparently a lot of H200s were stuck in Chinese customs for a while, spawning conspiracies about top-down directives.
People who don't regularly deal with Chinese customs may not realize that they are genuinely the single most corrupt, incompetent, and evil section of the Chinese government. Over the past 5 years, I have literally never heard about anything bribe-related that wasn't about Customs. They lost 4 seperate copies of my diploma during COVID. They will sometimes "lose" biological samples if not paid a bribe.
All a foreign adversary would need to do to delay hardware shipments to Chinese firms is to pay $500 over WeChat to some mid-level bureaucrat at Customs. Maybe that's what actually happened.
11I suggest investing ~100% of your money in developing-country funds. This is a generally good decision if you're young (high-risk, but also high EV), and seems much more sustainable than donating 10% of income. Maybe there's a related guide out there?
It's honestly my biggest gripe with the EA movement. I think a counterfactual EA movement which stressed more altruistically effective investment as the primary message instead of an actual tithe (40k/Catholic LARP does not generalize across cultures) would do far more good.
Working memory digit limits vary greatly based on language, averaging ~10 for Chinese (all digits are single short syllables) and ~5 for Hebrew (2 syllables per digit). For a conservative estimate of the upper bound, there's ~40 possible phonemes in human language, which can be extended to 160 if we use 4 different tones, and log2(160) is 7.3 bits. This implies that our auditory loop alone should be able to hold 73 bits of information based on easily pronounced phonemes alone.
I think you can push this much higher if you used a soundboard, which would free you from using only human-pronounceable sounds.
Currently looking at some of the example problems used to assess AI capabilities in this paper. This... obviously doesn't work as intended? Based only on recall (no search), Deepseek answers these questions somewhat well, getting 2/3 attempts for the first, 0/3 for the second, [assumes Star Wars refers to Episode I and gives the answer for that movie, changes to correct answer after clarification, so either 0/3 or 3/3]. It identifies Satan, but misidentifies the squished animal as a cat for #4 in all 3 tests. So a model with zero visual ability could score quite well just from the context given by the question and maybe the metadata of the upload file. The first question doesn't even require the name of the movie for a correct answer!
This seems to indicate fundemental problems with their test construction - it gives "written by someone from 1990 who makes an AGI test without ever interacting with an LLM" vibes. Maybe that was the intent?
They also cited actual College Board example AP tests from ~2014 as their sources for like 30% of their problems. I would be shocked they weren't in the training data multiple times.
This whole paper just feels wrong.
Talking about moral philosophy with your CCP boss in a genuine heart-to-heart, philosophical way is like, actually insane. Imagine reading a book where an American protagonist grabs a police officer's gun from behind and the officer laughing and saying "ahh, that's good old American CQB, my friend stranger, you won't get me next time!" and then the two becoming fast friends after shouting racial slurs at each other in a McDonald's for 20 minutes.
I would at least use after-work binge drinking to make it a bit more plausible.
1I was specifically refering to US import tariffs on consumer goods.
$300,000 is a big number, enough that it starts becoming cost-competitive for the government to just contract out every component of the gamete -> educated 18 year old pipeline.
That is, the state pays for donor gametes, IVF, surrogacy, and 18 years of foster care. There's a lot of confounding factors here, but the last time I ran the numbers it was something like $200,000-$300,000 per child.
At the very least, this is a solution that continues to work even if desired number of chidren craters all the way to 0 - daycares don't seem to have any trouble finding workers, after all.
I actually think modern urbanites have a lower effective social population density than the ancestral environment. Most hunter-gatherers have very little privacy, with entire entended families huddled in the same single-room dwellings. This level of density is very rarely present in developed cities (other than pubic transit). Most people in, say, Tokyo have their own room and thus few far less crowded than the average hunter gatherer, if you polled them throughout their day. Lowering the number of people encountered per day doesn't seem to greatly increase fertility, otherwise COVID would have resulting in a big fertility bump. Maybe the problem is the opposite - modern technology allow us to have fewer undesired social encounters, which is a high-priority desire for most people, but those same undesired encounters were the main force behind the formation of romantic relationships.
It is possible that the easiest way to increase the fertility rate is a legal mandate for dinner with co-workers.
I've never actually bribed a Customs official, so I don't know the exact process. I think they can initiate extra inspections, which can't be cancelled, but can be sped up?