An issue with this is that the average person doesn’t understand the AI labs’ position of “we think this might kill everyone, but we’re building it anyway because we think our competitors would be even more likely to kill everyone if they got there first”. Given that, saying “we think our product might cause human extinction” probably sounds bizarre or like hype. The linked post could explain this though.
carterteplica
Karma: 1
I agree with the general thrust of this, but posit that (for most people reading LessWrong) it is better to pick several hills to defend gently than one hill to die on.
I am atheist—I think it very unlikely that gods exist or that the supernatural claims of any religion are true. Nevertheless, I have basically positive affect toward religion; I grew up singing in choirs that met in the local Episcopal church, enjoy Scott Alexander’s writing on Judaism, think Buddhism has interesting psychological insights, and so on. I think it’s an advantage to me that I feel this way, because when someone asks me if I’m religious I can say “no” with a smile and move on. Similarly it comes up a lot in daily life that I’m vegetarian; it’s useful that I’m not viscerally disgusted by meat.
I think my non-rat friends’ impression of me is “nice and sensible, maybe has some weird philosophical investments with that EA thing”. Then, when persuading them of something (eg to vote for Alex Bores) matters, I can be like “as you know, I don’t usually try to go all in on persuading you of things, but here’s why I think this is unusually important and worth your attention”.
To some extent this take is just, like, “be agreeable in the Big 5 sense”; despite that I think a lot of aspects of this are pretty trainable.
I vibecoded a Thunderbird extension to do this: https://github.com/crtep/seriate. It’s cobbled-together but so far pretty good for clearing backlogs of mostly-junk emails.
Hm. I think this is trying to fit a wide range of interactions into a framework of
Initially, people have a variety of rights to discretion over actions of others (e.g., I have a right to decide whether or not people can have sex with me, take my possessions, or punch me).
Entering a contract means agreeing to consent to particular actions of others.
Another example might be: initially I have a right to keep you out of my house. If you rent a room then I permit you in advance to enter my house whenever you want (irrevocably, during some time and subject to some conditions). If I want to have you over for dinner then I permit you to enter my house only at one particular time.
As a matter of public policy, our modern society takes the position that one can give irrevocable advance permission only for some of these rights (entering my house or taking my objects but not having sex with me or punching me). There are some rights where I cannot permit you to do the thing at all (you killing me). Some rights can be transferred or delegated and others cannot.
What I think this misses is that typical contracts also involve agreements to perform physical actions, not just to waive or transfer rights. If I buy an apple from you then you agree not only to transfer your right to ownership of it to me but also to physically allow me to take it. If I buy a meal from your restaurant you agree to make it for me. If both of our labs agree to slow down work on AI capabilities then we create a new right (against violation of this) and also take some concrete actions (reallocating compute and employees, giving METR new API keys, etc).
I recently read and like Schelling’s treatment of agreements in the first few chapters of The Strategy of Conflict.