[Jach]

While I agree that 3 is the most optimal solution in the sense that "all these other problems are solved given 3", we still don't have a firm time frame on it. It could be 1 year, it could be 50 years. (I wouldn't put it at 100 or above personally, barring global catastrophe.) So we do things in parallel and hedge our bets. Could we get a self-sustaining colony on the Moon (or Mars, or somewhere else) within 50 years if we tried? I think we could. And that instantly protects modern humanity from many existential threats while we continue to work on problem #3.

[ansible]

Time frame is always a tough one. Let's try to bracket it with what we know, and be clear about the goal.

First, if we're talking about running an uploaded human-equivalent AI, we've got the processor power for that now, but it takes up a large server room. So I'd say we need to shrink stuff by at least 2 orders of magnitude to launch that into space. With corresponding gains in efficiency. So for that I think we're looking at 10 years at current rates of progress. Tack on another 5 for radiation hardening, because that estimate was based on commercial-grade hardware, which is almost as fragile as meat.

After you have that, it is a small matter of programming :-)