Win-wintegrity

Neuroscientist explains how the SAT is redefining education

This is the key moment between Jensen and Dwarkesh on export controls: 1. Dwarkesh asks why it’s okay to sell NVIDIA chips to China given the national security implications of AI models like Mythos. 2. Jensen gives a misleading answer, arguing that it’s okay to sell American chips to China because China already produces 60% of the world’s chips. 3. But as Jensen definitely knows, compute is measured in flops, not number of chips. 4. Dwarkesh then pushes back, pointing out that on a flops basis, China has 10% of the compute the US has, and giving them more compute would change their cyber capabilities. This exchange shows why it’s critically important for interviewers to have at least some technical knowledge, so they can push back against misleading talking points.

What’s impressive about this is that the education system didn’t just eliminate virtually all conservative thought as it moved toward left wing extremism, they got rid of almost anyone who was in the middle too. Anyone who doesn’t fully comply with the groupthink isn’t welcome.

@WhiteHouse yet another example of boomer communism on the backs of young people

@RosnySpanngen @AutismCapital No, just don't trade specifically threatening capabilities with the specific countries you are worried about posing a threat. Conditional cooperators survive.

Dumbest question ever though. Why is China supply the US military with rare earths or the “enriched uranium”? Why is the US providing Microsoft Windows, the real core tech the PLA is run on? Run this to its logical conclusion - there should be no trade because they may be a threat. Any country may be a threat. Your own country may be at civil war. Everyone produce their own food, clothes, and ai chips.

Loving Wartime Jensen

This is just straight forward evil. They didn't have enough kids to support themselves, and then they just vote to extract from the young. Completely unsustainable.

@danwilliamsphil Where does the instrumental convergence argument fail? If we get examples as we get more capabilities isn't that the opposite unfalsifiable armchair theory? I generally get annoyed with some doomer rhetoric, but they still have some good arguments.

Interesting post, though I see things a bit differently. The "AI people" typically bundled together two distinct ideas: (1) that extremely capable ("superintelligent") AI systems are possible, likely to be built eventually, and likely to create huge challenges that deserve far more attention; and (2) that the main challenge these systems pose is catastrophic misalignment, where superintelligent AI decides to wipe out or enslave humanity. Claim (1) has aged very well, and those who championed it deserve a lot of credit. That said, it isn't that intellectually impressive, because the writing was on the wall from the mid-twentieth century onwards — pretty much as soon as we figured out how to build computers and it became clear that intelligence is rooted in computation. Claim (2), on the other hand, hasn't aged well and was never plausible to begin with. It rests on dubious, low-quality armchair arguments (e.g., "instrumental convergence") and generally poor epistemics (see, e.g., mechanize.work/blog/unfalsifi…) Recent AI progress doesn’t, in my view, provide any real confirmation of (2). The "AI people" often point to things that allegedly confirm it — studies of LLMs lying and blackmailing, AI psychosis, MechaHitler, and so on — but at most these show that aligning and controlling AI systems perfectly is hard. They don't show that we will somehow build advanced AI that develops a motivation to wipe out humanity. For the most part, actual frontier AI systems are extremely well aligned with the intentions and preferences of their designers and users. To the extent that they pose serious problems, these overwhelmingly come from misuse, not misalignment (and perhaps also from the scale and pace of societal and economic disruption advanced AI will bring with it). So it's a mixed bag. It was impressive that the "AI people" flagged the importance of taking advanced AI seriously, but many of them misrepresented the risks. And I suspect that precisely because they bundled (1) and (2) together, many people's reasonable scepticism when they encountered (2) led them to discount (1) as well.

Quick update on the "autism epidemic": - Tylenol does not cause autism - Vaccines do not cause autism - Autism hasn't actually increased The entirety of the "rise" in autism is due to diagnostic drift rather than an increase in the population's liability to have autism.

Mental model: If there's a credential with a label attached, the less qualified people with the credential will identify more with the label. Ex: Less accomplished and less capable MDs, PhDs, and Ed.Ds will be more likely to demand you call them "doctor" or list their degrees.

The reason you see dorms or co-living arrangements only in college or the military is because colleges and the military have the ability to summarily evict a troublemaker without going through the 9 month eviction process ordeal that progressives have inflicted on the rest of society.

Harvard abolishes its SAT requirement in hopes of removing evidence of its preferential treatment of black and Hispanic students in admissions. A few years later, it creates its first remedial math class.

Among those who underwent medical "gender reassignment", the increase was stark. Feminising procedures: psychiatric morbidity rose from 9.8% to 60.7%. Masculinising procedures: from 21.6% to 54.5%. The procedures did not resolve the distress.

@cremieuxrecueil How many woke paths with there be first though?

We should make oaths more common and enforceable to promote trust. I like the idea of voluntarily subjecting yourself to higher standards in whatever domain, especially if that means you can be punished for oath-breaking. I want to apply this to all academics immediately.

Teenagers are sharing photos of their AP U.S. Government textbooks, and the sheer amount of indoctrination is wildly disturbing. Apparently, Barack Obama is ideologically a right wing authoritarian. Hillary Clinton and George W Bush are entirely indistinguishable politically. Donald Trump is of course virtually the same as Hitler. @tedcruz is apparently more radically authoritarian than Fidel Castro AND Joseph Stalin..??!!?? @Linda_McMahon — can we expedite some major changes to American public education?

It's not your imagination. SF is getting a lot better.

@Jakdrh @NavidFarda In terms of wisdom, earlier preemption of bad regimes also seems a lot better than letting a bad regime get to the point it could attack multiple countries and kill millions.

@Jakdrh @NavidFarda In WW2, England attacked Germany before Germany attacked England. Iran has been killing Americans for decades and the formative moments of the regime involved taking dozens of Americans hostage. Iran has tried to bomb nearly everyone in range, they're merely less competent.

این برنامه پیرز مورگان واقعا سورئال بود. دو تا ایرانی داشتند برای دنیا تعریف میکردند از دست چه رژیم خونخواری فرار کردن و یک چپول آمریکایی اصرار داشت که هیچ سندی در مورد جنایات رژیم وجود نداره. فکر کن یک انگلیسی مینشست جلوی دوتا یهودی آلمانی و میگفت: هیتلر خیلی هم براتون خوبه.

@JeffreyLuscombe Guess which state is building a Shinkansen. Guess which states have failed to build high speed rail over and over again. Conservatives just hate corruption.

Why do Conservatives hate this so much? Anyone?