One thing I find interesting: People who make weird-sounding predictions that later turn out true rarely get a lot of retrospective credit.
I think the psychology is something like: 1. Person A makes weird-sounding prediction that they genuinely believe in 2. Person B finds the claim weird and mentally saves Person A as “a weird person” 3. The prediction turns out to be highly accurate 4. Person B now believes the claim based on the external evidence visible to them (e.g. AI progress) but Person A is still stored as “a weird person” in their mind 5. Then if Person A makes a new prediction, Person B thinks “it’s the weirdos again” and, but since the external evidence is not available yet, they don’t believe in the object-level claim. 6. The loop repeats.
For example, I think @DKokotajlo & the AI futures project, have made some incredibly accurate predictions in 2021 about what 2026 looks like AND AI 2027 has turned out scarily accurate in a bunch of ways since publication.
And yet, a lot of reactions to the AI futures project predictions are like “it’s the sci-fi guy again, clearly WE would have predicted early 2026 to look like this if we had tried harder, but the rest of AI2027 is nonsense”.
