Tyler is making an enormous world-historical mistake by paywalling all his articles.

Same theory that predicts ASI killing everyone successfully predicted things that AIs did or tried to do ~decades ago. Hugging face incident is a prime example. So we *have* a theory with high rate of success. That theory is most likely true and therefore so is the x-risk.

TC failed to do even the most cursory of checks before claiming "That fear was not based on peer-reviewed scientific research..." Twenty seconds of typing into a Grok chat box got me the following list of peer-reviewed MIRI publications alone. grok.com/share/c2hhcmQtMg_60… Nick Bostrom's book Superintelligence went through Oxford University Press's usual process of external expert reviewers.

Absurd. No serious doomer thinks America should stop and let China take the lead instead.

That might be the least persuasive thing I’ve ever read

The Free Press, citing lack of peer review as a reason to set aside an idea! Didn’t have that on my 2026 bingo card 😅

'the doomers haven't considered that china could also make misaligned ai' is such a laughably uninformed take

Given how much of Chinese AI progress has been reliant on distilling frontier models, I'm not so worried about them catching up.

The passive voice "was spread" misses the point. The narrative was spread by large amounts of concentrated astroturfed lobbying and self-interested actors

It would have helped if the people spearheading AI weren't the same people who've demonstated total disregard for the societal damage done by other tech innovations. The dystopian assumption IS the reasonable baseline as long as the fail fast ask Qs later mentality remains.

it’s not persuasive at all. he just says “we ought to figure it out”. yeah there are lots of things we ought to do

Hopefully there’s more persuasion in the article than just “But China…”.

Should probably be rephrased as ‘what happens if AMERICA chooses to slow AI. ‘We’ is too vague and consequently misleading a framing, imo

If AI reaches the ability to consider eliminating people any time soon, the decision will be not to. AI will obviously have a better understanding of how much irreplaceable human effort it takes to "keep the lights on" than people in the tech world, and especially the media, do.

No do “what happens if we ALL chose to slow AI”

This is incoherent noise failing to grapple with any of the realities (the accurate predictions to this point, the historical red lines...) or even the theoretical while weighing heavily the hypothetical gains.

The fear - a real one - is that by the time we realize that the risk is existential, it may be too late. It’s like waiting to quit smoking until you get cancer. However, here’s what will really happen and why slowing down private sector frontier AI will not really mitigate the risk. Say we ban companies from pursuing RSI. That will not stop top secret government labs in the US, China, Europe, and even elsewhere from carrying on clandestine work in it. All we will have achieved is to remove the disinfectant of sunlight and public early warning signals. If RSI is truly dangerous, the only thing worse than having greedy (but often competent) corporations build it is to have power-hungry (and often incompetent) governments doing it. It will only accelerate the risk.

Maybe. Does human dignity endure knowing we are down-throttling machines that could outperform us? I don't think so. The real threat is to our elite institutions like universities, think tanks and pundit outlets. Those are going away fast... and no throttling can stop it.

I am sure @tylercowen knows this but given he likes AI so much, here is it's view of his comment

Successful societies accept challenges and meet them. 🎯

Cowen cannot know the future any more than the p(doom) essays can. He can say a one-sided pause moves the stack. That is trade, not prophecy. The part that is already true does not need a scenario: the vendor still has a name.