normative
Multiplying large by small numbers is unreliable
Long-horizon discounting calculations, such as effective altruists' estimates of AI risk many generations out, involve multiplying very large numbers by very small ones (or dividing by tiny numbers), so you can get almost any answer you want; the tools for thinking about such trade-offs are ill-suited because the magnitudes are entirely up in the air.
normativepending
Speaker
Suresh NaiduEvidence Quote
“thinking quantitatively about trade-offs at 10 generations from now, I'm not sure is super productive.”
Created: 6/13/2026, 12:09:56 AM
Your notes
Loading notes...