Excitement vs childishness

by sil ver 1 min read3rd Jan 2020No comments


I've heard Robin Hanson and others make the argument that people will be biased towards the fast takeoff scenario because it's exciting to think about (or "sexy"). On the other hand, there is a bias towards disbelieving the fast-takeoff scenario because it's childish.

I think most of us can agree that both are indeed biases, i.e. should be assigned zero weight, because both are about attributes that don't correlate with what's true. So we have the excitement-bias and the childishness-bias. The question is, how do they compare?

It feels totally obvious to me that the childishness bias is far stronger. I see people signaling maturity all the time, and childishness seems to be extremely low status in the relevant circles. It's so low status that it's not uncommon to see people say things about dangers from AI that are accurately summarized as "even though I don't know anything about this topic, I will evaluate its legitimacy based on how many childish sounding arguments I hear, because obviously they are not the real concern and people who defend them have zero credibility". Even among those who provide other arguments, it seems more common than not that they also assume that the fast takeoff scenario is less likely a priori because it's childish. Conversely, I've never heard anyone imply that fast takeoff must be true, or even likely, because it's exciting. It seems to me that, in order to have a similar effect, the excitement bias would have to do some rather heavy lifting in a purely subconscious way, which doesn't seem very plausible.

Despite this, I've seen far more discussion about the excitement-bias than the childishness-bias. That seems wrong.

Disclaimer: I don't think this is a strong argument that fast takeoff is likely, and I also don't think that bias towards excitement is weak in general – just that it's weak among the relevant class of people. Finally, I don't think this bias is common on LessWrong, only just about everywhere else.