Building mental habits requires some mental awareness to notice when they’re appropriate. How can people improve their mental awareness? Meditation might help because of detachment from thoughts, or maybe spending time self narrating and trying to notice every thought and feeling. It seems like a key skill not everyone is good at.
qazzquimby
Karma: 227
I’m thinking of artificial communities and trying to manufacture the benefits of normal human communities.
If you imagine yourself feeling encouraged by the opinions of an llm wrapper agent—how would that have been accomplished?
I’m getting stuck on creating respect and community status. It’s hard to see llms as an ingroup (with good reason).
Thanks for reporting this! Most likely it was because of ‘window height’ wasn’t excluding the parts covered by mobile browsers. I’m now specifically using ‘inner height’ which should fix it.
Wow I wish I had searched before beginning my own summary project.
The projects aren’t quite interchangeable though. Mine are significantly longer than these, but are intended to be acceptable replacements for the full text, for less patient readers.
Thank you, I hadn’t noticed the difference but I agree that complacency is not the message.
I think I can word things the way you are and spread a positive message.
Thanks a lot, you’ve un-stumped me.
I’m in the process of summarizing The Twelve Virtues of Rationality and don’t feel good about writing the portion on perfectionism
”...If perfection is impossible that is no excuse for not trying. Hold yourself to the highest standard you can imagine, and look for one still higher. Do not be content with the answer that is almost right; seek one that is exactly right.”
Sounds like destructive advice for a lot of people. I could add a personal disclaimer or adjust the tone away from “never feel satisfied” towards “don’t get complacent” though that’s a beyond what I feel a summarizer ought to do.
Similarly, the ‘argument’ virtue sounds like bad advice to take literally, unless tempered with a ‘shut up and be socially aware’ virtue.
I’d appreciate any perspective on this or what I should do.
In future should I post summaries individually, or grouped together like this?
Individual posts is more linkable and discoverable, but having a post for a full sequence of summaries might be more ergonomic to read and discuss.
Thanks for your thoughts, I’m glad I asked.
You’re right my goal isn’t very well defined yet. I’m mostly thinking along the lines of the https://non-trivial.org and https://ui.stampy.ai projects. I’d need a better understanding of beginner readers to communicate with them well. I’m not confident that I’ll write great summaries on the first try, but I imagine any serious issues can be solved with some feedback and iteration.
Would summarizing lesswrong writings to be more concise and beginner friendly be a valuable project? Several times I’ve wanted to introduce people to the ideas, but couldn’t expect them to actually get through the sequences (optimized for things other than concision).
Is lowering barrier to entry to rationality considered a good thing? It sounds intuitively good, but I could imagine concern of the techniques being misused, or benefit of some minimum barrier to entry.
Any failstates I should be concerned of? I anticipate shorter content is easier to immediately forget, giving an illusion of learning.
Thanks for your time. Please resist any impulse to tell me what you think I want to hear :)
I think that list covers the top priorities I can think of. I really loved the Embedded Agency illustrated guide (though to be honest it still leads to brain implosions and giving up for most people I’ve sent it to). I’d love to see more areas made more approachable that way.
Good point on avoiding duplication of effort.. I suppose most courses would correspond to a series of nodes in the wiki graph, but the course would want slightly different writing for flow between points, and maybe extended metaphors or related images.
I guess the size of typical Stampy cards has a lot to do with how much that kind of additional layering would be needed. Smaller cards are more reusable but may take more effort in gluing together cohesively.
Maybe it’d be beneficial to try to outline topics worth covering, kind of like a curriculum and course outlines. That might help learn things like how often the nodes form long chains or are densely linked.
Inspired by https://non-trivial.org, I logged in to ask if people thought a very-beginner-friendly course like that would be valuable for the alignment problem—then I saw Stampy. Is there room for both? Or maybe a recommended beginner path in Stampy styled similarly to non-trivial?
There’s a lot of great work going on.
I have an implementation here https://thirdtime.toren.dev
I guess it generalizes to: if there’s an unsolved problem and the solution looks obvious, you’re probably missing something.
Beware the natural tendency towards overconfidence
It’s easiest to think of the easy happy path. The problems are usually more nuanced and less mentally available.
Other people thinking about it are probably not being dumb or thoughtless.
I don’t think it’s caused me to overthink in that if something seems one dimensional it’s probably being underthought.
There are learnable exceptions, like a friend might have a mental blindspot to a certain kind of solution, or you might consistently overthink certain situations.
To be honest, I’m still not a great listener because I haven’t squashed the urge to think of advice before empathizing.
Thank you for making my floundering into something actionable...
I’ll first try looking into what people have found before on this thinking. I find it surprisingly difficult to see what my outgroups are or what advice I should be thinking of reversing.
I’m pretty sure it would have been easy to find ten more of most of those, but it would have felt like cheating.
I felt the same way. It’s easy to generate something similar to an existing choice, like I included both catapult and trebuchet, but it feels wrong. But when I think about it feeling wrong, that’s premature pruning...
find the moon’s postal code and mail
catapult
shot-put style
by waterslide—to the moon
hack the package’s location value
bring the moon to the package
create a children’s book “how the package reached the moon” with a choose your own ending page, and use the most promising submission
space escalator
photoshop the package on the moon
trebuchet—superior to the catapult
just build a bridge
look at the moon, hold up the package, close one eye, and position it just right
close my eyes and imagine the package is on the moon, then leave the room
name your house “the moon”
email the moon announcing their package has been delivered, and ignore any responses
disguise the package as a baseball and hope
just hope
rocket
hot air balloon rising in escape velocity
moon portal (remember airlock)
pully system
E.T. bike
the package is already at the moon
ask the moon to come pick it up from a local post office
jump on a see-saw with the package on the other side
tie it to a homing pigeon in a homing pigeon space suit
put it on a boat, and then raise the water level
jump down from the moon, onto a trampoline, and carry the package on the way back up
pledge 50$ to whoever delivers the package to the moon
cannon
cannon, thrown by a trebuchet, on a rocket
smuggle it in past moon customs
give the package to a sea turtle
attach a curb to the moon and olley up it
tie to balloon and consider job done
put it on the moon-train
flubber
remove the earth and let gravity do it
stack boxes to climb up
up strong the package from the top platform
bet someone they can’t get the package to the moon. Call them chicken if necessary.
wait for the moon to fall into the earth
get a full 3d scan of the particles in the package, have it digitized and reproduced on the moon
flick it with a giant hand
roll the package into an extremely long and thin baton and just hand it over
go to the past and sneak it into the shuttle
go to the past and sneak it into astronaut food
cast “Deliver to Moon”
put the package on the moon’s christmas list
ball and chain flail around the earth
offer the package to the moon as a purely symbolic gesture
find a place on earth called moon and bring it there
Thank you! I had been looking through tags, and even thinking “what I really need are ‘techniques’”—yet I did not search for techniques.
The “drilling down along a new and different branch of the tree” concept makes me think of tree search algorithms, naively being depth or breadth first searching. It’s overly simplified, but might uncover related theory.
The goal is to search from whichever node you estimate to being closest to the goal. Calculating the estimate is difficult, so we tend to only look at a small nearby neighbourhood, which is usually low level. Backtracking forces you to make estimates for earlier nodes.
If I was making this algorithm faster, I’d try to find a way to make the heuristic (the estimate of nearness to the goal) more efficient. I’ve no idea how to do that, but maybe looking at how past discoveries were made could help.
Then again, given that research takes a long time, maybe it’s not worth making any sacrifices to the heuristic accuracy.
I’m not certain if this qualifies as a planning fallacy, but I’ve noticed a class or problem where a large nebulous task isn’t made actionable, and we just expect it to happen at some point. More an error of “when it will be done by” than “how long it will take.”
For example, my family knew for maybe a year that we would benefit from an exercise machine, and had discussed it many times. It was only when I realized the problem and set a deadline for myself that we actually got it.
I’d be very interested in talking to anonymous friend, or anyone else working on this. I have two relevant projects.
Most directly, I wrote a harness for llms to play text adventures and have spent some time optimizing the wrapper and testing on anchorhead. As you’d expect it has the same issues, but cheaper and without vision problems.
I’ve also worked on llm social deduction game play, which is nuanced and challenging in different ways, but shares the need for strong memory and robust reasoning in the face of hallucination.
I’d be happy to talk about any of these issues and compare leads!