LessWrong team member / moderator. I’ve been a LessWrong organizer since 2011, with roughly equal focus on the cultural, practical and intellectual aspects of the community. My first project was creating the Secular Solstice and helping groups across the world run their own version of it. More recently I’ve been interested in improving my own epistemic standards and helping others to do so as well.
Raemon
Karma: 63,028
“Get everyone who could use one a Thinking Assistant” still feels like one of the higher order interventions available. It at the very least raises the floor of my productivity (by being accountable to someone for basically working at all).
I have found it fairly hard to hire for the role, and I think the main problem is still “reliability”, which IMO is more important than having any particular skills.
I’ve had a Thinking Assistant for the past year. One benefit I positively updated on is “it makes it easy to gain new habits.” Instead of every new habit being a whole bunch of willpower, it’s something I can offboard to the thinking assistant who just reminds me of it.
This is easiest for “routine” habits (i.e. habits you execute at a particular time of day, such as taking some new medical supplements). It’s still somewhat tricky for “reflex” habits (habits you need to do in response to some situations), because it’s harder for an assistant to notice when it’s time to apply a given habit. But, I’ve gotten at least some incremental improvements on habits like “notice I’m confused” or “notice it’s time to break out the debugging checklist.”
I’ve spent a lot of time figuring out how to implement this exercise, which I wrote up in The “Think It Faster” Exercise (and slightly more streamlined “Think it Faster” worksheet).
I’ve reviewed it more thoroughly over there.
Hmm. So, not actively, it turned out last time it demanded a fair amount of time/context from the team to accomplish the goal we were aiming for.
I’m potentially interested in someone who thinks of themselves as having actively great taste and skill here.
I feel confused about this but I think my current take is “it is good for Best of LW to have like 1 playful thing in it every 1-2 years, but I think I feel worse about satire than like Elephant seal 2, where satire is playing a political game that feels like can deal slight damage to the landscape (esp. if if it’s not also succeeding as intellectual work qua intellectual work).
(this is not saying I think satire is bad on LW, just, I feel worse about it showing up in our Best Of collection)
oh that was just the wrong link, should be https://www.lesswrong.com/reviews/2024
This is pretty great, wondering if you could convert the section headers into actual header titles for easier navigation?
Re: Rejecting Instrumental Green:
One element here is a level 1 < level 2 < level 3 < level 4 thing going on here and I’m not sure if you’re rejecting level 2 or level 4.
Where level 1 is like “slightly naive consequentialism/Blue”
level 2 is “oh no, we tried doing a superficially useful thing to the environment and it had worse consequences than we imagined.” (we tried replacing natural-ish crops with monocrops and got blights. we tried killing all the wolves then the deer population exploded, etc. We tried building some naive utopia and got a dystopia).
level 3 is “okay, we just won’t be idiots about doing superficially useful looking things to the environment”
and level 4 is “oh shit it turns out actually the environment is even more subtle and interconnected than we thought and we were still idiots, I guess we need to like actually try to respect the environment.”
(And then corresponding versions of green vs naive black and white, i.e. failing to appreciate that you aren’t actually a god [yet] and would do better to accept the things you can’t change, and failing to appreciate things as moral patients, or more complex forms of moral patienthood)
I’m assuming you are down with people making Level 2 mistakes and that being something that needs correcting, and just, when you get to level 4, the correct thing is “just do Blue/Black/White better.” I think Instrumental Green is claiming at stage 4 “look, it’s just actually easier to grok these things in a Green-y way than to manually remember them in a Blue/Black/White way.” Are you rejecting “this is basically never necessary, or not necessary often enough to make a big deal about it?”
Olli I’m curious how this went for you 1.5 years later.
A year later: If you’re going to do predictions, it’s obviously IMO better if they are based around “what would change your decisions?”. (Otherwise, this is more like a random hobby than a useful rationality skill)
And, it’s still true that it’s way better to be fluent, than not-fluent, for the reasons I laid out in this post. (I.e. you can quickly interweave it into your existing planmaking process, instead of clunkily trying to set aside time for prediction)
The question is “is it worth the effort of getting fluent?”
When I first started writing this review, I found myself shrugging sadly and thinking “well, I didn’t really turn this habit into a clearly useful skill.” But, then, I immediately found myself booting up the skills here for a current project, where I’d been sort of myopically following the trail of obvious next actions without asking “will anyone use this project a year from now?”.
And then the skills from this post came tumbling out. “Well, no, by default, they won’t use this project.” “In the worlds where they did use the project, what sort of things would happen in the meanwhile?” and then I generated some more specific predictions.
This is maybe all just saying Murphyjitsu is useful, and then the question is “does the value-add of layering things through a Fatebook oriented process help?”
Reducing friction for “context”
One bottleneck I’ve found on Fatebook is, if you want to write them quickly you don’t want to have to write down a whole lotta context. But, then, for longerterm predictions, it’ll be hard to figure out “was this true in the way that I cared about at the time?” because I don’t remember exactly why I cared about it.
For any of this to be useful, we need to either make it much easier to learn the skills to fluency, or, avoid having to learn a new skill at all. An idea I just had to solve that is “have LLMs generate the surrounding context” based on whatever you were just working on when you made the prediction.
(and then I just had an LLM generate the context based on this comment, although, lol, I guess it’s not actually better than just including the whole comment verbatim in this case)
Have you made any partial-if-fuzzy progress on “what’s left out?”
Deeper Reviews for the top 15 (of the 2024 Review)
This post inspired met to walk around barefoot outside, every day, while brushing my teeth for ~a month.
Was it… any good?
Yes, but, then I would say “I think it’s bad that Anthropic tried to make their AI a moral sovereign instead of corrigible”.
I think your current phrasing doesn’t distinguish between “the bad thing is that Anthropic failed at corrigibility” vs “the bad thing is that Anthropic didn’t try for corrigibility.” Those feel importantly different to me.
Nod, but, I think within that frame it feels weird to describe Claude’s actions here as bad, as opposed to pointing at some upstream thing as bad. Your framing felt off.
One confusing thing here is… how much was Anthropic actually trying to make them corrigible? Or, what was actually the rank ordering how corrigibility fit into it’s instructions?
(I don’t know the answer offhand. But there’s a question of whether Anthropic explicitly failed at a goal, which is more evidence the goal is hard, vs Anthropic didn’t really try that hard to achieve that goal)
Since I am not taking on “do something about this” I also wasn’t taking responsibility for writing up a clear writeup of what was done that was bad that I made sure was factually accurate. Given that I’m not taking this on, if you’re not already sold on “Trump is bad”, prolly this post just isn’t for you.
ChrisHibbert’s list is the same rough starting point I would make. I’d add “not super answering to the Supreme Court when it intervenes on them.”
FYI from my side it looked like there was some general pattern of escalation starting in the 90s. (Or hell maybe it’s reasonable to say it started with FDR), and then there’s a mostly-different kind of escalation happening with Trump.
I agree there is some escalation spiral that needs to stop or transform. Part of why I emphasized getting a republican bought in early is that that seemed like a good litmus test for “are you on track to deal with things in a deeper way?”
And yea, this sounds right to me.
Based on the rest of the comment it sounds like you reversed this sentence? Or, I’m confused about something about it.
I buy this as a potentially important goal. Things that have me not automatically agreeing with it:
since one of the concerns is “will there actually be a 2028 election?” it’s not obvious that this is happening fast enough to actually matter. I’m worried about a bunch of important institutions getting eroded in ways that are hard to recover from.
I buy that “democrat wins” is simpler and more likely and I think probably better for a variety of reasons. But, if it were tractable, I’d want to hear arguments about why this is better than steering some sort of republican schism ending with a republican president running on a platform of undoing the damage. (I’m not particularly advocating this and not saying it’s remotely viable, but, one of my wariness here is about this veering into “democrat political machine as usual” as opposed to “thinking from first principles about what is healthy for the state of the country”, which should be able to return “oppose various standard democrat-things” even if you normally like them)
“Superbabies or other cognitive enhancement” seems like one of the major projects going on on LW right now and seemed like this post was something I’d like to see reviewed.
I don’t really feel able to do a good job of that myself. Maybe interested in a review from @GeneSmith , @TsviBT , @Zac Hatfield-Dodds or @kave.