Effective Altruists: look, we tried to invite nice people as well as the huge racists we knew were huge racists when we invited them. What? Exclude the racists? But they're so interesting!
I said in a comment the other day that Roko’s view here is the natural consequence of Yudkowsky’s original naive physical-scientific reductionism. He proceeds from those (abysmally vague but superficially straightforward) premises here. In essence, if everything ultimately reduces to the physical, then when you perform the natural reduction on e.g. the status of black people in modern America, the causes must by physical-biological causes.
The reference to Lysenkoism is perfectly apt on this (stupid) model: attempts to thwart the reduction are merely ideologically driven cludges to the real theory, and the example of Lysenko demonstrates how easy it is for a whole discipline (in this case: biology in the USSR) to fall to that ideology. Liberal (read: communist) biologists are just pandering and making exceptions when they produce their own demonstrations that scientific racism is bunk.
It’s helpful that, for historical and political reasons, i.e. America and modern Europe’s original sins (colonialism and slavery) scientific racism is always waiting in the wings when the Rokos of the world reach their inevitable conclusions. Put it’s important not to conceive of scientific racism as a form of ignorance: it is, rather, an often highly organised political movement devoted to proving and promoting its claims by any means necessary - it is a knowing lie, with the caveat that insofar as scientific racists frequently show that they implicitly know that they lie (with absurd clandestine promotion strategies and revealing statistical sleights-of-hand), it’s rarely clear that they are wholeheartedly aware of it.
Roko is also violating their rules of assuming charitably and good faith about everything and going meta whenever possible. Because defending racists and racism is fine, as long as your tone is careful enough and you go up a layer of meta to avoid discussing the object level claims.