📈🚨New preprint on graph perception🚨📈
We all know that some plots and charts are easier to grasp than others. But do we all share a minimum level of 'graphics' intuitions? And how can we measure it? We tried to answer these questions here! A thread🧵1/9
biorxiv.org/content/10.110…

My main fear is that EA will do none of these things, and instead will be consumed by sci-fi longtermism. The billionaire money will be captured by mediocre AI startups and liberal cyberwar think tanks, producing little value in either the short or the long term.

My main wish for EA is to abandon the fundamentalist subjective Bayesianism it inherited from online rationalism, to diversify its community and eject the “human biodiversity” cranks, and to find common cause with a broader range of scholars

This all drives me insane bc I think EA is good but needs strong criticism, and I disagree with both the critics and the EA core

The public discussion is about people and ideas I care about, but is so polarized that it’s unlikely to change anything. Depressing! Anxiety-inducing!

The allegations that MacAskill is a eugenicist are unfair, but the community does have people who think there are racial, genetic differences in IQ that justify paternalism (though they often leave this as subtext in public), and the acceptance of these people is deeply worrying

This radical conceptual openness and the tendency to bite bullets also opens the community to people who have been pushed out of the mainstream because they have terrible ideas

It’s easy for a non-Bayesian to see how this kind of community could go off the rails, but very hard for subjective Bayesians in the community to see it because Bayesianism is part of their identity and many are contrarians that feel they are courageously biting a bullet

So EA solicits criticism, but in practice only accepts criticism in a subjective Bayesian frame. Arguments that try to change the conceptual possibility space are reinterpreted as to merely shift credences a bit, leaving longtermist expected value calculations pinned to infinity

Another role subjective Bayesianism plays is limiting the effect of criticism: Arguments that the possibility space is ill-defined or incoherent are often rejected because they’re not Bayesian enough

One role subjective Bayesianism plays here is focusing discussions on forecasting the likelihood of AI risk rather than the conceptual possibility of AI risk. As someone who suspects “superintelligence” is basically conceptually unsound, this rings all kinds of alarm bells

E.g., many in the community think that even if the chance of a superintelligent AI killing all of humanity is very low, we should invest massive resources to prevent it, as on a Bayesian account the expected value approaches infinity. This has become an longtermist EA cause area.

Particularly so because when presented with consequences of subjective Bayesianism that seem perverse, community members tend to bite the bullet

So as someone who has spent a little time studying philosophy of science, the exclusive focus on subjective Bayesianism is really strange. It feels religious!

This is out of step with reality. There are plenty of epistemological systems that are at least as “rational,” including flavors of falsificationism, objective Bayesianism, social epistemologies, various hybrids

One of the downsides is that it encourages fundamentalism: I often see the attitude in the community that subjective Bayesianism is the One True Way, the only possible rational truth-seeking procedure

I think game-like structuring of this Bayesian spiritual/self-help content tends to draw in people that are a little socially isolated and unsure of themselves and willing to integrate Bayesianism into their identity in a deep way. There are upsides and downsides to this!

I think online rationalism draws people in through a network of densely interlinked texts and forums designed to produce experiences of Bayesian insight

The experience of reading these texts and forums has the flavor of an Augmented Reality Game or “doing your own research”