Opens profile photo
Follow
Click to Follow RichardMCNgo
Richard Ngo
@RichardMCNgo
What would we need to understand in order to design an amazing future? Figuring that out
San Francisco, CAthinkingcomplete.comJoined October 2009

Richard Ngo’s Tweets

I do think that even one example of massive failure suffices to rebut simple versions of "competent benevolent elites" narratives. So covid policies, housing dysfunction, etc should be big updates for many! Just not big enough to swing them all the way to the other pole.
3
12
Show this thread
Pending a better term, I'll call it inadequacy spiraling, since it involves getting sucked in to narratives about civilizational inadequacy. Inadequacy spiraling is less bad then Gell-Mann amnesia though, because impact is heavy-tailed: one big success pays off many mistakes.
1
13
Show this thread
We need a term for the opposite of Gell-Mann amnesia, where a few civilizational failures (e.g. covid) make you think the whole world is arbitrarily crazy. Blanket heuristics are rarely helpful; the alpha is in figuring out where the world is craziest, and fixing that.
4
64
Show this thread
This is the good stuff. It's only 3k words but there are 3 or maybe 4 separate plot twists I totally didn't see coming. (link in next tweet)
Quote Tweet
wrote about the mortifying ordeal of being known (MOOB-K), but also aliens, and somehow drunk driving again
Show this thread
Image
1
6
Show this thread
A whistleblower says that US partners of the Wuhan Institute of Virology said the experiments to put furin cleavage sites into SARS-like viruses (as close as you can get to a recipe to create the Covid-19 virus) were done in 2019.
35
874
Show this thread
But more broadly, as moderns, we think about the past through the framework of elites because through the miracle of modern technology we live like elites. But that's the point, isn't it? Tear that down and you are back to being a peasant. You do that, I'll pass. /end
37
2,174
Show this thread
The best therapists are genuinely interested in all the different aspects of what you tell them, even when you're working through complicated issues, and that makes such a difference. Or in other words: I'm long the long tail who long to detail long tales.
Quote Tweet
hard to articulate how radical the delta between a pretty good therapist and an amazing therapist is
Show this thread
1
27
London Underground workers earn twice as much as the national median wage despite the technology to replace them being solved decades ago. Worth keeping in mind the unions and regulation will have a big impact on the pace of change
60
178
Image
Quote Tweet
Image
How often have you heard it said that people have higher life satisfaction as they age, along w/ data purporting to support this notion? It turns out it's not true. Most data is cross-sectional, which is biased b/c happier people live longer, on average. prb.org/resources/happ
Show this thread
8
307
Many people, including me, have been surprised by recent developments in machine learning. To be less surprised in the future, we should make and discuss specific projections about future models. In this spirit, I predict properties of models in 2030:
22
545
Show this thread
Described to someone at a party tonight as "the best place in the world to meet people who deeply understand therapeutic and relational practices in analytical and systematizing ways". That's why I'm going, at least. Curious if this resonates with others though.
4
58
This is of course a very standard argument for libertarianism: that while government control often helps in *theory*, it's almost always done badly in practice. In particular, the mental health costs of lockdowns (esp to young people who were least at risk) were way underrated.
4
31
Show this thread
In hindsight, my ranking of covid lockdown policies (from best to worst): 1. Limited well-targeted lockdowns (e.g. of crowded indoor spaces, excluding outdoor gatherings) 2. No lockdowns 3. The lockdowns that actually happened (too harsh and protracted, aimed at the wrong things)
5
50
Show this thread
Three concerns about Georgist land taxes: - If building new homes remains incredibly hard, developers would pay high tax for the long period between buying land and getting permission to build - I expect that in most cases land taxes are politically infeasible unless they have… Show more
4
7
New huge twin study of brain structure, cognitive ability, & psychopathology confirms the 3 Laws of Behavioral Genetics: Everything is (partly) heritable; Families matter much less (often not at all); Lots of random influences (confusingly called 'unique environment').… Show more
731
We’ve released a statement on the risk of extinction from AI. Signatories include: - Three Turing Award winners - Authors of the standard textbooks on AI/DL/RL - CEOs and Execs from OpenAI, Microsoft, Google, Google DeepMind, Anthropic - Many more
174
1,221
Discrete optimization is difficult. Choosing the ideal number of children to have is easy - it's picking an integer number of children that's hard.
5
46
Square profile picture
With more powerful AI systems comes more responsibility to identify novel capabilities in models. 🔍 Our new research looks at evaluating future 𝘦𝘹𝘵𝘳𝘦𝘮𝘦 risks, which may cause harm through misuse or misalignment. Here’s a snapshot of the work. 🧵
32
730
Show this thread
More on abundance: theatlantic.com/ideas/archive/ It may appear that there's a tension between this position and my advocacy for extensive AI regulation. But there's not: I think we should have more regulation of tech with potential to cause catastrophic damage, and less of other tech.
3
15
Show this thread
The first step of cultural change is naming what you're for and what you're against. I'm for creating widespread abundance via people being able to get stuff done. I'm against vetocracy. Examples: NIMBYism, most occupational licensing, FDA overreach, etc.
Quote Tweet
I went to my first town meeting last night and my conclusion is that the "America can't build physical stuff" issue is fractal. Which suggests that the issue is cultural more than the result of any specific <federal, state, local> law. twitter.com/YIMBY_Princeto…
Show this thread
5
85
Show this thread
If we hadn't had wars already economists would look at their model of the world and conclude they shouldn't occur because they're so inefficient. And they'd be right, most of humanity isn't at war most of the time, yet wars are still very important.
Quote Tweet
Economists seem to consistently be the most dismissive of AI existential risk concerns, out of all groups of people who think seriously about the future. Why is this and what can we learn from it? twitter.com/DAcemogluMIT/s
Show this thread
11
151
Show this thread
something like an IAEA for advanced AI is worth considering, and the shape of the tech may make it feasible: openai.com/blog/governanc (and to make this harder to willfully misinterpret: it's important that any such regulation not constrain AI below a high capability threshold)
224
1,494