Sequences

Random Attempts at Apllied Rationality
Using Credence Calibration for Everything
NLP and other Self-Improvement
The Grueling Subject
Medical Paradigms

Comments

I don't think the "US censorship community" had any problem banning things shared on TikTok. They could just say "If you don't ban the Bin Ladin letter discussion, we take it as a reason to invigorate our banning efforts".

Other important companies that are affected:

Riot Games is 100% owned by Tencent.

Tencent has a 40% stake in Epic Games. Given that Epic Games is working on their own app store, being able to pressure them to censor apps is likely of interest.

Eliezer wrote The 5-Second Level. 

If you identify a bad 5-Second step, 30 seconds give you six training runs where you can go through the step to train it.

One core problem with AI is that it's not just "people" who make up new things behind teh scenes but AI itself that will make up new things. 

TikTok continues to be Chinese spyware. It also continues to be an increasing point of vulnerability for China to put its thumb on American culture, politics and opinion.


As long as it's legal for anyone to be a data broker and sell out a lot of data about the American population to anyone willing to pay the price, which includes China, shutting it down for being spyware makes little sense.

If you truly care about data not reaching China you need GDPR-type data privacy legislation and enforce that legislation. 

One, continued demands that we ‘prove’ that China or ByteDance has its finger on the scale. I say that the information elsewhere in this post, in the absence of counterexamples, is very strong evidence. At minimum, they are fixing content balance issues they dislike and allowing those they like, with an algorithm that snowballs.

TikTok banned discussion of Bin Ladin's letter not because the CCP disliked it but because the US national security establishment disliked the letter. The same is true for them banning Glenn Greenwald. 

While they might change their content moderation policies in the future in the moment those decisions are made to appear nonthreatening. 

I am sympathetic to ‘rushed’ objections when bills are so long there is no time to read and understand them. This does not seem to be one of those cases.

If you think you understand the implications, which other popular social media company is partly under Chinese control and thus targeted by the bill if the government would argue that it violates national security concerns?

Tencent owns around 38% of Discord's shares.

This law makes it easy to pressure Discord to do whatever the White House wants them to do by threatening to ban them otherwise.

Transposons activity is not downregulated in the placenta while being downregulated in most other cells. If there are too many transposons in the DNA that likely makes the placenta fail during the pregnancy. As a result, the amount of variance in transposon counts you see when sequencing born people is more limited than that of embryos.

If you do IVF you care about the pregnancy not terminating after three or four months. 

The evidence is more of a theoretical argument. In the selfish gene frame, transposons (even when they aren't technically genes) "benefit" from copying themselves even when it reduces the total fitness of the organism.

You need a mechanism that prevents people with too many transposons from procreating to account for transposons counts not just growing indefinitely.  

In the absence of a process that regulates a quantity, you have variance. Lack of observed variance in sequenced adult genomes is evidence of a mechanism for regulation existing. 

By the mid-2010s, you could genotype all the parts of a person’s DNA most likely to differ from other people’s for under $100. At that price point, it became possible to gather genomes from hundreds of thousands of people and assemble them into giant databases that researchers could access. 


I think there's a good chance that transposon count in the DNA differs between people, shotgun sequencing can't tell you how many times a given sequence appears.  

There's a tradition in biology to consider things that can't be easily measured as irrelevant, but that doesn't mean it's true.

If you have a nonstandard unicode character in your password and an attacker tries to crack a password based on the assumption that no nonstandard unicode character is in the password, they can't crack your password no matter how much compute they throw at it.

Given that you decide where in your string you place it, the attacker would have to test for the special character being at multiple different positions which adds a lot of additional entropy.

Load More