If I recall my past opinions correctly, I said at the time that while such wars were the only way to free certain countries, I did not trust the competence of the current administration to prosecute it and was strongly against the way in which it was carried out in defiance of international law.

I would say in retrospect that the resulting disaster would have been 2/3 of the way to my reasonable upper bound for disastrousness, but the full degree to which e.g. the Bush Defense department was ignoring the Bush State department was surprising and would not become known until years later. I have since adjusted my political cynicism upward, and continue to argue with various community-members about whether the US government can be expected to execute elaborate correct actions based on amazingly accurate theories about AI which they got from university professors (answer: no).

less than 4,000,000 slaves

Depends on how long slavery would have lasted without the war. You don't just free current slaves, you prevent future generations from ever being enslaved. I take that to be Yudkowsky's point- the earlier the better, because the more slaves you prevent.

Edit: Actually, use your own numbers:

a slave could only expect to suffer for 20 years on average while a free man typically lived to 41

This implies converting someone from slave to free should increase their life by 20+ years, approximately half the life time of a free person. Just multiplying, using 1 free person dead = 41 years lost, one slave freed = 21 years gained, we pass a cost benefit test.

Backing out a little... if I see two people about to shoot someone tied to a tree, on your view am I ever justified in shooting them both to save the tied-up person's life? If so, what does it minimally take to justify that?

"Whenever I hear anyone arguing for slavery, I feel a strong impulse to see it tried on him personally.

-- Abraham Lincoln

in other words, the suffering of a lifetime of slavery is evidently worth 18.75% of the death of a free person in a horrific war.

Leaving aside all moral considerations of collective responsibility and individual complicity... and switching to my rough model of preference utilitarianism, which I generally don't use... this would sound like an incredible, unbelievably lucky bargain with this cruel universe at HALF a life for a freed slave. At 18,75% it appears perverse even to hesitate in this non-dilemma.

P.S.: instead of preference utilitarianism, I do find it much more comfortable to use broadly Christian virtue ethics for a snap moral decision. According to which... well, let's just mention that even a Catholic like Chesterton could be unapologetic in his respect for the Jacobins. Never mind the Christian abolitionists of the day.