Cruxes

What should be done about AI? Most disagreements about that question are really disagreements about a few factual questions underneath. A crux is a question where a different answer would change what someone thinks we should do.

Each page makes the strongest case on each side, links to the arguments and sources behind each case, and says what evidence would change minds. None declares a winner; a post can argue for an answer.

Can it be made safe?

• How hard is alignment — The central question for most strategy disputes.
• Can we learn from AI failures before they are fatal — Whether iteration works, or the first serious failure is the last.
• Can AI automate alignment research — Whether a short window can be enough.

Can it be slowed?

• Can AI progress be paused — The forms a pause could take, and whether any could hold.
• Can the US and China agree to limit AI — The geopolitical version of the pause question.

Given those answers, what should careful people do?

• Should safety-focused labs keep building — Build carefully, or stop and warn.

If it goes right

• Who should control aligned AI — The power question that remains after a technical success.

The SURA debate turns on how doomed Plan C is shows these cruxes at work in a single debate.

The answers depend on one another. If alignment is tractable and automatable, staying in the race looks more reasonable. If alignment is very hard, the pause questions become the important ones.