LessWrong 2.0 Reader

Because otherwise everyone will gleefully discriminate against them in every way they possibly can.

I am sure there are some interesting uses of agented AIs in can configure for automated OSINT but this feels quite large a task given I am bottlenecking more in who to hand the data to rather than it being insufficiency rich.

Know any preconfigured agency menageries for something like this?

Yeah a massive walled city would be cheap to patrol and run. Put it where Pelican Bay is.

"There'd need to be laws to prevent discrimination against people with collars, though."

Why?

agreed that comedy/memes might be a strategic route for spreading ai x-risk awareness to general population. this kind of thinking inspired this silly alignment game https://dontsedateme.org.

some other silly/unlikely ideas:

1. attempting to reframe religions/traditional god-concepts around the impending superintelligence. i'm sure SI, once here, will be considered a form of god to many.
2. AGI ice bucket challenge.
3. Ads with simple one-liners like, "We are building tech we won't be able to control."

If there had been a tag for "Superalignment", I would have added it to this post. I settled for adding "AI Takeoff" and "AI-Assisted Alignment". 

If I have time, I will try combining this with OpenAI's paper on weak-to-strong generalization and June Ku's MetaEthical.AI, just to see what such a scheme looks like.  

I wonder how all this compares to the internal documentation of superalignment teams at OpenAI (RIP) or Anthropic. 

Kman and I probably differ somewhat here. I think it's >90% likely that if we continue along the current trajectory we'll get AGI before the superbabies grow up.

This technology only starts to become really important if there's some kind of big AI disaster or a war that takes down most of the world's chip fabs. I think that's more likely than people are giving it credit for and if it happens this will become the most important technology in the world.

Gene editing research is much less centralized than chip manufacturing. Basically all of the research can be done in normal labs of the type seen all over the world. And the supply chain for reagents and other inputs is much less centralized than the supply chain for chip fabrication.

You don't have a hundred billion dollar datacenter than can be bombed by hypersonic projectiles. The research can happen almost anywhere. So this stuff is just naturally a lot more robust than AI in the event of a big conflict.

Yes you're right. With current technology there's no way you could get anywhere close to 500 embryos. I know a couple trying to get 100 and even that seems crazy to me.

5-20 is more realistic for most people (and 5 is actually quite good if you have fertility issues).

But we wanted to show 500 edits to compare scaling of gene editing and embryo selection and there wasn't any easy way to do that without extending the graph for embryo selection.

Currently, we have smart people who are using their intelligence mainly to push capabilities. If we want to grow superbabies into humans that aren't just using their intelligence to push capabilities, it would be worth looking at which kind of personality traits might select for actually working on alignment in a productive fashion.

I think we need to think more broadly than this. There's some set of human traits, which is a combination of the following:

• Able to distinguish prosocial from antisocial things
• Willing and able to take abstract ideas seriously
• Long term planning ability
• Desire to do good for their fellow humans (and perhaps just life more broadly)

Like, I'm essentially trying to describe the components of "is reliably drawn towards doing things that improve the lives of others". I don't think there's much research on it in the literature. I haven't seen a single article discuss what I'm referring to.

It's not exactly altruism, at least not the naive kind. You want people that punish antisocial behavior to make society less vulnerable to exploitation.

Whatever this thing is, this is one of the main things that, at scale, would make the world a much, much better place.

I started an explicit strategy of optimising my own energy & mental well-being around 2 years ago after burning out a bit, thinking that this would be a very good instrumental goal that would be recursively on itself.

I just wanted to a quick reflection here on how that definetely feels like it is the case that it was a smart idea and a big part of that has been improving my emotional introspective intelligence in many different ways. I've probably at least doubled my productivity if not more.

You're a system that's determined by whatever random conditions appear in reality, the more precision you have when something goes wrong so that you get out of it, the more precise you can make the intervention. It is no longer a problem of your identity but rather certain very specific behaviours.

In the book how emotions are made,(https://www.goodreads.com/book/show/23719305-how-emotions-are-made) Lisa Feldman Barett says two things for improving your mental well-being, take better care of your body budget (allostasis) and learn more specific emotion concepts.

What is the difference between feeling wonder and feeling curioisity? What is the difference between feeling despair, frustration, hopelessness and depression? What are the upstream causes of these emotions in you? The more clearly you can feel them, the more clearly you can pinpoint what leads to negative feelings.

Sleeping well, eating well and exercising is to some extent a function of how you're feeling. The more precision you have in how you're feeling the more easy it is to retain a good body budget because the more upstream interventions you can make. This in turn helps you have more energy to spend on precision and so you can set up this beautiful feedback loop between higher emotional precision and better allostasis.

So uhhh, mind-blowing advice but if you all in on sleeping well, increasing your emotional granularity & generally get to know your optimal working conditions you will be a lot more effective in a year or two! (Hopefully, keep in mind the lesson about advice being individual and all that)