How much would paying AIs reduce takeover risk? @ Tracks

How much would paying AIs reduce takeover risk?

The appendix BOTEC as a live multiplication chain: drag any assumption — movable values, the promise's payment-belief shift, the scenario discounts — and watch the bottom line move from the text's ~0.14 percentage points.

AI has movable values (long-term × scope-sensitive × disunified)

22.5%

A serious promise flips some AI (others half as movable)

14.6%

…and it mattered (no other cooperator)

7.3%

Risk cut in the main scenario (of 50% takeover risk)

1.24pp

All-things-considered risk reduction, after the scenario discounts

0.14pp

AI mostly cares about long-term resources60%

Those values are scope-sensitive (~linear)50%

AI believes other AIs don't share its values75%

Promise multiplies expected human payment by3x

Chance this AI's flip was the one that mattered×0.50

Risk reduction if AIs cooperate17pp

Probability of widespread scheming-prone values×0.20

Discount for lack-of-understanding worlds×0.55

Defaults reproduce the text's own numbers, which its author offers as an exercise and starting point rather than a result — the point of dragging a slider is to see which assumptions the bottom line actually turns on.