Comments
Note: This post was crossposted from Planned Obsolescence by the Forum team, with the author's permission. The author may not see or respond to comments on this post.
Subtitle: If acceleration compounds, the naive extrapolation implies superintelligence within a year
This post discusses a public METR report on AI R&D acceler...
With your aggressive tone, it's perhaps understandable why you've run into mod trouble on LessWrong. But as a simple existence proof, the forecasting techniques and training materials described by Phillip Tetlock in books like "Superforcasting" have been repeatedly shown to somewhat improve people's skill at making all kinds of predictions across varied subject areas. Forecasting isn't the same thing as LessWrong-style "rationality", but it's close -- both are general reasoning skills that focus on avoiding bias and understanding probability, rather than domain-specific expertise.
https://80000hours.org/problem-profiles/improving-institutional-decision-making/#there-are-several-ways-to-make-progress-on-improving-decision-making