Hey uh, I’ve been thinking all of those thoughts too. We should probably nucleate up a community (a slack channel or something, somewhere to hang out and share our findings and make plans) because I’m pretty sure there are at least 10 people knocking around just here who have their heads as far into this as we do. Heard Eliezer was absolutely overflowing with discursive technologies when Arbital was being planned, his concepts were fractaline. I’ve been that way. I guess I pulled back a bit when I started to understand that having infinite visions of sophisticated collective intelligence augmentation systems isn’t really the hard part, the hard part is building any of it, funding it and holding users.
I do see some ways to do those parts.
I’m just gonna start talking excitedly about the most recent piece of the puzzle I turned up because until this moment I have not had many people to talk to about this (lots of friends who’re interested but not many who’d ever take what I was saying and do anything with it)
Yesterday I flipped out a little when I remembered that article Scott Aaronson did about eigenmorality (eigentrust) and I realised it this is exactly the algorithm that I’ve been looking for months, for doing a basic implementation of the thing you’re calling “Contrast Voting”… (I’m going to keep calling it order-voting and graph rank recovery if you don’t mind? Idk I think there are more standard terms than that) I haven’t tried it yet (I just found it yesterday. Also I want to port it to Rust) but I’m pretty sure it’ll do it. Basically what we need to implement order voting is, we need a way of taking a whole lot of order pair judgements/partial rankings from different users and combining them together into a single global ranking. With eigentrust (similar to all the other stuff I’ve been trying), basically what we’ll do is we’ll build a network graph where each edge represents the sum of user judgements over the two candidates, then we run eigentrust on the graph (it’s a similar technology to pagerank, if you’ve ever had that explained to you. Score flows along directed links until an equilibrium is reached), and then we have an overall ranking of the candidates. We’ll need to do some special extra stuff to allow the system to notice when two clusters of candidates aren’t really comparable with the data we have, and it’ll probably need to try to recognise voting cabals because there’s a dark horse problem where-...
I should really write this out properly somewhere.
The reason I haven’t done that already is that I’m not sure how many of our concepts should be exposed publicly.
these technologies are actually powerful. Even just order voting alone would speed up content sorting by like 20x, imgur could use it for recommending fucking cat pictures and they would become even more compulsive than they already are. (They might already be using it in a hidden way, I think netflix is.) Power isn’t good or bad on its own, but some powers are more likely to be put to good uses than bad. Collective intelligence platforms are more likely to be put towards good uses than AGI is, they’re inherently made of humans, they’re more likely to reflect roughly human values even when they go wrong, but in their worst incarnations they can still just end up becoming completely insane demonic egregores like… dare I even speak their names, no, no I daren’t, because I don’t want to draw their millions of eyes towards me. Let’s just say that some of the social media platforms I frequent most often are basically incapable of forming sound epistemic structures, and I’m afraid of most of their segments there, and I really hope that those words they’re saying never become much more than words.
The ideas I have for technologies that’d gather and harmonise users quickly and efficiently are also some of the ones that scare me the most. I know how to summon an egregore, but Making the egregore come out of the portal sane takes a special extra step. It’s absolutely doable, but I wouldn’t trust anyone who’s not at least weirdsun adjacent to understand the problems well enough and to stop and think about what they’re doing and put in the work to make it all turn out human, and maybe not release it onto the internet before it’s sound.
I think the first step is to make something that gathers information that people want. A place where people will feel comfortable forming communities and spending time. A humane place, something that respects peoples’ attention, rewards it.
The world needs platforms where good mass discourses can exist, currently we have, actually none.
I actually think this should be an EA cause. At some point, if we can gather a decent team, we should start asking for funding. Maybe move to the EA hotel in blackpool and grind on it for a bit once we have a 1.0 vision.
I...only followed some of what you said here. *Googles slack channel* … Sure, if you know other people who are interested in a similar concept, that might be worthwhile. How do we go about it?
I’m very excited about what might happen if we got ten people like us in a channel, I think that’s a community/project I’d give a lot of energy to, but that didn’t occur to me until just partway through reading your post, so I have not been collecting any names until this point, sorry. Maybe we should wait til we have a few more than two, before I start sending out invites (by the time we do, there might be something nicer for async group chats than slack).
(weirdsuns are… analytic surrealists. I don’t know if I’d say they’re influential, but as a name for a certain kind of thinker, those unmoored by their artificial logics from the complacency of common sense, they’re a good anchor on which to ground a label.)
Hey uh, I’ve been thinking all of those thoughts too. We should probably nucleate up a community (a slack channel or something, somewhere to hang out and share our findings and make plans) because I’m pretty sure there are at least 10 people knocking around just here who have their heads as far into this as we do. Heard Eliezer was absolutely overflowing with discursive technologies when Arbital was being planned, his concepts were fractaline. I’ve been that way. I guess I pulled back a bit when I started to understand that having infinite visions of sophisticated collective intelligence augmentation systems isn’t really the hard part, the hard part is building any of it, funding it and holding users.
I do see some ways to do those parts.
I’m just gonna start talking excitedly about the most recent piece of the puzzle I turned up because until this moment I have not had many people to talk to about this (lots of friends who’re interested but not many who’d ever take what I was saying and do anything with it)
Yesterday I flipped out a little when I remembered that article Scott Aaronson did about eigenmorality (eigentrust) and I realised it this is exactly the algorithm that I’ve been looking for months, for doing a basic implementation of the thing you’re calling “Contrast Voting”… (I’m going to keep calling it order-voting and graph rank recovery if you don’t mind? Idk I think there are more standard terms than that) I haven’t tried it yet (I just found it yesterday. Also I want to port it to Rust) but I’m pretty sure it’ll do it. Basically what we need to implement order voting is, we need a way of taking a whole lot of order pair judgements/partial rankings from different users and combining them together into a single global ranking. With eigentrust (similar to all the other stuff I’ve been trying), basically what we’ll do is we’ll build a network graph where each edge represents the sum of user judgements over the two candidates, then we run eigentrust on the graph (it’s a similar technology to pagerank, if you’ve ever had that explained to you. Score flows along directed links until an equilibrium is reached), and then we have an overall ranking of the candidates. We’ll need to do some special extra stuff to allow the system to notice when two clusters of candidates aren’t really comparable with the data we have, and it’ll probably need to try to recognise voting cabals because there’s a dark horse problem where-...
I should really write this out properly somewhere.
The reason I haven’t done that already is that I’m not sure how many of our concepts should be exposed publicly.
these technologies are actually powerful. Even just order voting alone would speed up content sorting by like 20x, imgur could use it for recommending fucking cat pictures and they would become even more compulsive than they already are. (They might already be using it in a hidden way, I think netflix is.) Power isn’t good or bad on its own, but some powers are more likely to be put to good uses than bad. Collective intelligence platforms are more likely to be put towards good uses than AGI is, they’re inherently made of humans, they’re more likely to reflect roughly human values even when they go wrong, but in their worst incarnations they can still just end up becoming completely insane demonic egregores like… dare I even speak their names, no, no I daren’t, because I don’t want to draw their millions of eyes towards me. Let’s just say that some of the social media platforms I frequent most often are basically incapable of forming sound epistemic structures, and I’m afraid of most of their segments there, and I really hope that those words they’re saying never become much more than words.
The ideas I have for technologies that’d gather and harmonise users quickly and efficiently are also some of the ones that scare me the most. I know how to summon an egregore, but Making the egregore come out of the portal sane takes a special extra step. It’s absolutely doable, but I wouldn’t trust anyone who’s not at least weirdsun adjacent to understand the problems well enough and to stop and think about what they’re doing and put in the work to make it all turn out human, and maybe not release it onto the internet before it’s sound.
I think the first step is to make something that gathers information that people want. A place where people will feel comfortable forming communities and spending time. A humane place, something that respects peoples’ attention, rewards it.
The world needs platforms where good mass discourses can exist, currently we have, actually none.
I actually think this should be an EA cause. At some point, if we can gather a decent team, we should start asking for funding. Maybe move to the EA hotel in blackpool and grind on it for a bit once we have a 1.0 vision.
I...only followed some of what you said here. *Googles slack channel* … Sure, if you know other people who are interested in a similar concept, that might be worthwhile. How do we go about it?
What is weirdsun?
I’m very excited about what might happen if we got ten people like us in a channel, I think that’s a community/project I’d give a lot of energy to, but that didn’t occur to me until just partway through reading your post, so I have not been collecting any names until this point, sorry. Maybe we should wait til we have a few more than two, before I start sending out invites (by the time we do, there might be something nicer for async group chats than slack).
(weirdsuns are… analytic surrealists. I don’t know if I’d say they’re influential, but as a name for a certain kind of thinker, those unmoored by their artificial logics from the complacency of common sense, they’re a good anchor on which to ground a label.)