David James comments on AGI safety from first principles: Conclusion

[David James]

Personally, I am most confident in 1, then 4, then 3, then 2 (in each case conditional on all the previous claims)

Oops. A previous version of this comment was wrong, so I edited it. The author’s confidence can be written as:

$P\left(1\right)\ge P(4|3,2,1)\ge P(3|2,1)\ge P\left(2|1\right)$

$P\left(1\right)\ge P\left(4|3\right)\ge P\left(3|2\right)\ge P\left(2|1\right)$

Also, independent of the author’s confidence:

$P\left(1\right)\ge P\left(2\right)\ge P\left(3\right)\ge P\left(4\right)$