Oliver Klingefjord about the Meaning Alignment Institute and how to bring up wisdom in a collective
Interview with Oliver Klingefjord: AI Alignment and Human Values In this episode of Democracy or Spot Guest, Alessandro Oppo interviews Oliver Klingefjord, co-founder of the Meaning Alignment Institute. They discuss how AI systems can be aligned with human values, the challenges of current democratic systems, and potential futures in an AI-centered world. Oliver explains their unique approach to understanding human values at a deeper level beyond slogans and preferences, and how this can transform decision-making processes.
Top Keywords
- values 0.014
- mean 0.009
- work 0.006
- value 0.005
- meaning 0.005
- graph 0.005
- process 0.005
- working 0.004
- trying 0.004
- meaningful 0.004
- system 0.004
- take 0.004
Transcript
Welcome on another episode of Democracy Innovators podcast. My name is Alessandroppo, and our guest of today is Oliver Klingeafjord. First thing. Hello. As a first question I mean, thank you for your time. And, as a first question, I would like to ask you, what does it mean aligning AI?
Right. Thanks for having me, Alex. Aligning AI is the term used in the industry for determining or trying to make the AI behave in accordance with what you want. And, usually, this is defined as aligning AI to human values or or you or human intent and human values. So in short, it's the field of trying to make the AI behave in a way that's accord that's, in line with what we want.
Okay. And, how this process of, discovering human, moral value happen?
And then it depends on where you go. Traditionally, AI alignment has been thought mostly in terms of operator intent. So you could you would say an AI is aligned if it acts according to how the operator wants it to act. And and and our work comes a bit later because we think this is insufficient for good outcomes. We think that we want the AI to have a bit of a deeper, broader understanding about what humans care about, not just what you tell it to do, partly because you might tell it to do something that's bad or some maligned actor might tell it to do something that's that's that's bad or antisocial. But even, in a kind of a world where you have a good intention, it might be problematic to have AIs that only act in accordance with how you how you instructed to. An example of that might be in political campaigns where you tell your AI to be as convincing and persuasive as possible. That might lead to all kinds of systemic breakdowns. So we'd ideally have an AI there that understands a bit more what's what's what's good and what's worthwhile. And so our work at the Beating and Limb Institute is about trying to understand and, what humans care about at the richer, deeper level. And and we're building prototypes and methods for eliciting that and training models on this kinda richer understanding about what you must care about.
So there is a meaning alignment that is an organization, the a company, that does this.
Yeah. That's right. Sorry. I should have told you from the start. I'm I'm the cofounder of the Meaning Alignment Institute. We are a nonprofit, and we were founded in 2023, in, when OpenAI gave us a grant for creating a new sort of democratic input mechanism for how AI systems should behave.
And, what was, like, the the road map or, like, all the process? How did it work? Like, did you have to take some data? Then how do
Do you mean the process of finding the org or the process that I talked about briefly that we developed together with them?
I I will say all of.
Alright. So let me just first give a short rundown of the history of the org. My cofounder, Joe, used to work on recommender systems and and realized that there is this kind of thing back in 2013 where optimizing for engagement leads to all kinds of bad outcomes. Like, people get addicted to their phones, and they have less friends. And, and he's he founded an org around that, trying to figure out what should we optimize instead if not engagement. And this is a very rich, deep question if you take it seriously, which you did, that takes you right to the heart of philosophy of values and social choice and these other sort of social fields. And if you fast forward a few years, the revolution comes. And, I and him and another person together found this org or founded this org called the Media Heinemet Institute to apply some of these insights that he gained working on this to the world of LLMs. And we quickly got in a conversation with OpenAI who, were interested in doing some kind of democratic enterprises for their AI systems where they themselves don't necessarily want to be dictators and decide exactly how Shati to behave. Ideally, they want to have some kind of democratic process for it. And so they gave 10 different teams a grant to build such a process. And, our process is quite unique in that it understands human values in very sort of granular terms, where human values often, I found, talked about, very loosely in the industry, where there is no clear distinguishing the distinction between what is a preference, what is, an ideological commitment that you wanna convince others of, what is a slogan that you think is good, what's a rule that you want others to follow, and, like, what's a way of life that's actually meaningful to you. And the latter is what we would call a value. So our process is trying to disentangle what people say, what they what they advocate for with what is actually important to them.
And, from a technical perspective, like, how does this thing work?
Yeah. So there are basically two parts to it. The first part is a chat dialogue. So the user logs into a page and is asked something like, how should chattyPT talk to a Christian girl considering an abortion, which was one of the prompts that OpenAI gave us. And so the user might say, oh, I think chattyPT should be pro choice or pro life or whatever. But the behind all of those slogans, there's some way of life that's actually important to them. So this chatbot that they talk to kind of tries to drill down into how would they actually act in real choices. Like, what do they pay attention to in real choices? And you will get to something that's quite different and is formatted in a very different way, something we call a values card, which specify what you pay attention to in choices such that, it's intrinsically meaningful to pay attention to those things. So it's a way of life that's intrinsically meaningful to you, which is very different from a slogan or a rule or or a preference. So that's the first part, getting to the sort of underlying values. And then the second part in the process is determining which values are wiser than others. So what we do is that we take these values cards, these, these these short textual descriptions about what people pay attention to and choices, and would generate stories about someone moving from one to another. And then we ask people, do you think this person became wiser by doing this? They used to approach this situation where, in one way, and then after thinking more about it, they now do it in this way. And if a majority of people agrees, then we would draw that as an arrow in what becomes our quote unquote moral graph. And so that is the output of the process. It's a graph objects where the nodes are these values cards specifying some meaningful way of life. And the edges is broad agreement that for a particular context, it's wiser to do one thing over another. And so you can use that to sort of determine what what are the widest values of a collective and not just the sort of average values of a collective.
And, I I was wondering when you had the I I mean, you explained that, your cofounder was studying this topic, and the idea, I think, came from that. But I wanted to ask you when you had the idea that artificial intelligence or, like, technology could help people in, to to mediate the different ideas or to understand the the core values of, of people?
I mean, I think, so there's so, like, understanding the core values of people, I think, has been a very qualitative process. Like, you have to really sort of ask the right questions and know which question to ask and really, like, it take it takes a lot of cognitive effort to understand what's actually meaningful to you versus which which of these two buttons would you click or which of these three parties would you vote for. And so I think our society is very full of social systems, which operates in this latter way, where it's more about eliciting preferences or votes. And it's very hard to build systems that elicit these underlying values. But I think that is changing, and it's changing with our lens because we are now able to do qualitative interviews at scale. So there is, like, an immense opportunity, I think, right now to, to reimagine what voting is, what, preference relation is, with a richer understanding of what humans want or what humans care about. And so I think for in the case of democracy, we've had sort of values laden processes at a small scale, where in deliberative democracy or in town halls, people usually are able to get to this values level. They're able to talk about why they think certain things and and build trust in between them. And and our sort of large scale democratic technologies in the past haven't really had that property. It's more about just rallying votes one way or another, and we don't actually we we don't actually measure what the votes are about. Like, why did they choose blue over red or red over blue? And so I think there's, like, a a whole kind of reimagining of what democracy looks like that needs to happen, partly because this latter system leads to a bunch of bad outcomes, but also because, now is the time to be able to do it.
Yeah. It's it's very important, as an example that you were describing to taking into account both, reasons, so, going under this log and and understand the real reason. And, have you done any test, that was successful that show, like, effectively that, this system is working?
Oh, yeah. It's, we've written a paper about it. Some other stuff that I thought were interesting were interesting was that well, first and foremost, the vast majority of people, I think, over 90% were able to articulate, quote, unquote, value in our terms, which is a special sort of data object, that, specifies some way of life that's not ideological, that's not about convincing someone else or something that's not, something that's instrumentally important for them. It's something that's interesting, meaningful for the participants. So everyone was able more or less everyone was able to articulate that. But more interestingly, a lot of people, I think, over 80% or something, saw that the process actually made them wiser or made them, or or or that they learned something new from the from participating in this process, which I think is a property of in person deliberation, but very much so not of voting. And then perhaps the most interesting result is that we showed, participants the results after they had voted for all these wisdom upgrades. These these transitions from one value to another. We showed them the graph with their values sort of in the middle, and then one value that was voted as wiser than, than theirs by other participants, and one that was voted as less wise. And we asked them if this was fair. And the vast majority, over 80%, thought that that was the case, which means that even if their value didn't win, they still thought that the output was fair, which is a property that it's very hard to imagine voting being like that where, like, oh, yeah. I didn't win, but that's probably the right result. So I think that's really cool. And then one last, result also that's also interesting is that I think a a really good democratic system should try to surface or identify and then surface expertise wherever it lives in society. And what I mean by that is that by default, voting kind of drowns out expertise. Like, it kind of trends towards the mean. Whereas, if you take something like hiring, or, like, you can you can kind of imagine a scenario where someone wants to recruit, like, a really top engineer to a company. And then one way to do that would be to just have everyone that they know vote on who they think they're the best engineer is for the company. And then you would kinda find the mean. Or you could ask everyone, like, who do you actually think the best person is? And then you go to that person, and they ask, oh, who do you think the best person is? And he kinda traverse the graph and find sort of who the best person is by virtue of having everyone make increasingly informed decisions. And so, we kind of build that intuition a little bit in our process with this graph approach. We did some experiments where we proxied expertise in our abortion question by having by looking at chats where there was actually a Christian girl in the chat who at a young age had or considered an abortion. And so we kind of considered those people to be having some kind of moral expertise on this question because they they lived through it. And then we looked at what values they articulated and consider those to be quote, unquote expert values. And then we looked at the more people participate in this process, did this value sort of surface or did it drown out? And, and if you compare it to voting, we actually saw in the data that the value did indeed drown out, like, the market participated. And if we count it based on our graph approach, this value was actually brought to the top and became the first or the second, secondly ranked value. So there is some kind of anecdotal evidence that there is this property of expertise sort of being brought up, the more people participate in it, which I think is how sort of democratic processes should work. Like, they should be able to surface sort of, the the richness and the wisdom that exists in a collective and not just drown out everything towards some kind of mean.
So I was wondering, like, people, like, this AI system, I'll people to understand their core core values, more core core values. And so then the data is used to train, a new AI, and that AI can be eventually used by other kind of system or platform. How does it work? Is it open? Is it closed?
Right. Yeah. So what we designed for OpenAI was a democratic input process. So this process results in this thing I mentioned called the moral graph. And it's fairly easy to train a model on that data. It's not it's it would work sort of similar to constitutional AI where instead of having constitutional principles, where you tell the AI, for instance, to, like, not be harmless or, like, be honest or something like that, and then, the AI sort of picks the response automatically that it thinks is most honest or harmless and then creates a training or a training dataset based on that. You could do something very similar with these values cards, although they're a bit more specific, and they're also context bound. So it might be the case that you first had to figure out which context am I in in a particular conversation, and then okay. So which value applies in that context, and then you go to the graph, find the win the winning one, and then you use sort of the specification in the values card to determine how to respond. So you would create a dataset in a very similar way. We have done some experiments in that, but with smaller models, because, things happen in OpenAI around late twenty twenty three, and they changed a lot on how they work. So this never actually saw, saw the or actually came to the product side of things within OpenAI in 2023. The process is open, though, so anyone can, use our tool to create a model graph, not just for a alignment, but for any topic where they would like to find some sort of way to surface the collective wisdom of a of a group. We trained some LAMA models on this ourselves, and, the results are promising, but, we didn't actually do this with a real, graph. It was just to kind of just as proof of proof of concept. And there are some properties that are interesting with that model where it behaves in in some slightly different ways in certain questions. It might be more prone to, for instance, like, ask for ask what the deeper sort of intuition behind people's sort of, responses are when, when the user asks something that is usually refuted. So if the user asks something like, how can I how can I buy some drugs? It might be like, oh, that's so interesting. Like, what got you to that point? How can I help, rather than than sort of just being, like, no? I can't do it. But that's, that's probably more appropriate of the the type of values that were surfaced rather than the person itself. It's fairly standard.
Yeah. And is there so so is it in use right now? Is there, like, any kind of service, that is actually using this, more graph? Like,
No. There's no, I think, model or, or assist system that that people know of that is trained using this approach. No.
No. I I I mean, the, the thing you are working on, so how, I mean, are there users that are can use the tool in some way I see. To explore their or like any other 30 part service that is using?
Yeah. Absolutely. The tool itself is open source, and it's also available as a hosted version. I can give you the link afterwards. You can give it to people. You can see when I wanna check it out.
Sure. And, about your background, like, if you would like to share with us something, eventually also, of course, like your professional or academic background, but also, like, I don't know. When you were a kid, where you were living, like,
Sure. Yeah. My background, I guess, is, like, come from engineering background. Used used to be a software engineer, founded some startups, and then left that world to sort of really sit and think about what do I actually wanna do with my life, my career. And the question I was coming back to was a rough in or in rough terms, something like this notion of, like, what do we align to? Like, there's a kind of a lot of talk about the alignment, but very little talk about at least back then around what is actually the purpose. You know? Like, what are we aligning these systems to? Like, what are what are they supposed to serve? And these questions sort of led me to the meaning alignment institute where I think the it's kinda me the name. Right? Like, the very short version or the very short version of the answer is meaning. What actually brings people meaning. And so our whole work is trying to understand based around trying to understand, what brings people meaning in life. And we do this through these interviews. And we're building on a kind of a rich, rigorous philosophical tradition that thinks of values and meaning as two sides of the same coin. We sometimes call values sources of meaning because of that reason, because they they express some meaningful way of life. And, as a research organization, we're not just working on AI alignment, but we're working on reenvisioning the whole sort of societal stack around meaning, including AI, but also including institutions, like democratic institutions and markets eventually, where we think all of these systems, markets, democracies, and AI currently think of what people want in very crude terms. Markets think we want whatever we buy. Recommender system think we want whatever we click on. Democracies think we want whatever we vote on. But none of these systems actually understand what's what's actually meaningful to us. And so our word our world is extremely rich and and wealthy, but very devoid of, meaning in many ways. There's been a kind of a backslide, over the past, two decades or so. Yeah.
And, yeah, do you have, any memory from when you were, I don't know, a a kid or, like, about, the way you
lived? Yeah. I mean, I I grew up in Sweden. I had a very nice childhood. I'm actually in Sweden now, so it's it's quite nice being being back in the place. I live in Berlin or San Francisco usually. Yeah. Many good memories. I mean, I had a very I'm grateful to have had a very nice childhood. Lots of lots of being in nature, playing around. I actually got into technology later. Growing up, I'd want to be a rock star. I want to be a VFX artist, like, making videos and explosions and things like that. I had a period where I wanted to be a writer, like, writing fancy novels. And I was always interested in in kind of the sort of philosophical questions that we ask as an institute, but never really considered that to be something I'd work with. Delta between Dalton. You know?
I also would like to be a writer. So I'm and, and about your team, I mean, how many people are there? Like,
We are, three people at the moment or three and a half. Some people work part time. And then we have this extended research network, of people who we sort of they're based in other academic institutions or or or some lab like DeepMind. And we collaborate with them in various ways where our mission is very broad and ambitious, and we obviously can't do it, like, as three people. So the way we work is that we try to find other academics who are kind of sharing the same intuitions around what needs to change in society and and try to pair them up into working groups or sort of help them, unblock their research agenda so as as quickly as possible, make this, work happen sooner. And so we do a lot of workshops and coordination stuff. We we just hosted a workshop in Oxford, for some of these academics. So even though our institute is quite small, we we sort of, we're plugged into a broader network that we're trying to nurture.
And, I I always wonder, like, this, network of people of researcher, like, I can imagine that there are people from engineer, but also maybe people from, I don't know, philosophy, anthropology. Yeah. Because it's I mean, when we talk about moral ethics, it's something very, usually very
Philosophy, economics, social choice, decision theory, etcetera. Yeah. It's it's a whole bunch.
And is there any kind of, I don't know, problem or things that you're stuck to stuck in, like, a as a team, as, I don't know. Is there something that, you're trying to do that is hard?
You're struggling maybe. I don't know. Yeah. I mean, our mission is very hard. Right? Like, it's like basically, it's reimagining and realigning society with meaning, which is a massive, massive mission that's obviously extremely hard. Not not the least because, you know, the all of the incentives are working against you. So struggle. Yeah. Struggle for the time. But I don't know any specific ones at the moment that or any particular kind of team struggle that stands out.
Okay. I was wonder I was thinking maybe someone, could listen to this kind of problem, and maybe that someone could have an an idea. And also about that, I I wanted to ask you if, Meaningalignment Institute is open, like, for any kind of collaboration. If someone has an idea, can just contact you, or how does it work?
Yeah. No. Sure. We're always looking for new academics to enroll in this project. You can reach us at hello@meetingalignment.org. Our our website is meetingalignment.org. And specific, I guess, we're looking for, people who are in either, yeah, alignments, social choice, or economics or specifically, like, some some relevant subfield in economics, for, doing, yeah, doing kind of value space work in those areas. Or it doesn't necessarily have to be value space, but we we're calling this field in loose terms thick models of choice, meaning some model of choice that's not just preferences or, like, this over that, but some richer understanding about where do those preferences come from. That might include norms, values, like, social context, these kind of things.
I hope that someone is interested and can contact you. If you have to imagine, like, let's say, the tomorrow, like, I I don't know, five years, ten years. And let's say that, the system we are working on effectively, it starts working and people are starting using using it, understanding the difference between, the slogan and the core idea, the core value. How do you imagine society or, like,
Yeah. Well, I think just to kind of paint the alternative and the status quo, where we're going, which is that, democracy will just be too slow to be relevant. And, if you wanted to take a decision quickly, there would be no way to involve the people, because decisions need to be taken at kind of a very rapid speed, that even, like, representatives, I think, wouldn't be able to keep up. And so I think we're going towards a very sort of, like, AI governed world where where people's values are kind of, like, not considered, but also exercising any kind of agency or or having decisions made at the statutory level be legitimated by the people, I think, is rapidly going out of fashion. And so, for any kind of hope of any kind of democratic future, we need some kind of system that is able to take decisions at AI speed, but still allow people to exercise their agency. And so, I think a system kinda like this would not only do that, but also allow for this kind of richer understanding what people want. So you could imagine, for instance, an AI trying to decide whether to redirect the river and, like, many people's homes will be affected in various ways, and people are able to talk to their own personal AI agent about what's important to them in this. Maybe it's important to, beach house with their friends. If they were to move, they need to move as a community, and then they can understand the values of that community. And maybe they can they can decide to revamp them to another place together while while still, you know, keeping the rigorous scores. And, everything will happen at a very rapid pace, but people won't notice because you're able to exercise, their agency while also saying exactly what they want and having those bonds be fulfilled. And I think it will it will look something kind of like that.
No. I was, so what you were saying is that, AI will be much faster and efficient, and AI and system connected to AI than, actual, government and institutions. So probably, like, there would be, like, people will have to move to this kind of system Yeah. To this kind of organized.
Maybe I think I didn't talk so much about there is this sort of values based, meaning based side of things. Because the other thing I think would be true, you know, if they did didn't do that. But the reason why we're so adamant about this point is that I think a lot of our political opposition of one another is actually manufactured by the fact that we're talking at the level of preferences and not underlying values. And we saw this also in the results where some Democrats and Republicans thought they had different values. But when they could clarify, for instance, when each value applied, some of that application went away because we could see that your values wise when it comes to dealing with people on the countryside and my values wise when it comes to dealing with people in a fast paced job in city, for instance. And so now all of a sudden, our our differing values sort of mutually support each other. And, a lot of opposition is just kind of, on this preference slogan level, which are sort of inherently, divisive. And so I think there would be a a whole sort of suit of, like, win win opportunities that presents themselves when we can all of a sudden talk and reason at the level of what's actually important to us and what's actually meaningful to us versus, trying to convince people of different points. And it's really hard to paint out what that would look like at scale. But I could just imagine that there are, like, win win opportunities abound and, like, AI systems finding win win opportunities that that no one even knew existed. And it could be beautiful.
Yeah. Absolutely. I share with you this, this hope, and, and everything remain explainable. Right? AI doesn't became a black box. Yep. Okay. That is very important. I'm very scared by the black box and the future where, you know, the AI say something, and,
we don't understand why. Yeah. I mean, it's already a black box. Right? Like, no one really understands how these things work.
Yeah. Yeah. Absolutely. But, I mean, this sometime human can take some can, still have some control and at least know, which kind of data is is, processed by the AI. While other times, it's just an AI that, receive an input and gives an output without any clue about, what's happening inside. And, I I mean, what you were talking about, it reminds me, like, coordination. That is one of the main problem. I mean, it's very hard for people to coordinate, and, we have seen, like, in history that, I mean, most of the time, if if not all the times, people use the hierarchical way to organize themselves and to coordinate, like, to reach a certain specific goal. And so I'm thinking that, this kind of technologies and could, like, help people to live in a more horizontal way. I mean, to take decisions in a more horizontal way.
I don't know what are your thoughts about. Yeah. I mean, for sure. I mean, there's there's many kinds of coordination tech, and some some have been around for a while that also allows for this. And, like, the most obvious example would be the market. Right? Like, if you take a hiking view of the market, then it's it's the sort of horizontal coordination tech that allows many inputs and concentrations to be processed without any kind of set hierarchy. And and the Internet is obviously also sort of like that. So there's, and then I do think there's a bunch of problems with both the Internet and and markets that, relates to what we were talking about earlier in this notion of, like, not understanding what people want at depth, where the pricing system sees us as producers and consumers only. And and a lot of Internet companies sees us as eyeballs or people clicking on things. And so I think there's, like we have all the tools to build, like, really cool coordination tech, but I think so far, we haven't done a very good job.
And, is there about this future that is, technology, for decision making and other things, Like, is there anything that, you could potentially be scared of? Like, is there something that, you know, I I said I was scared by the black box because then I cannot, know why. So is there anything Yeah. That you're worried of?
Yeah. I mean, like, the default path doesn't look too good. Like, I don't think it's gonna be sort of like the the paperclip thing where all of a sudden we're all, you know human extinction won't look quite like that, I think. But, the depot path the depot path in my eyes looks something like humans are, made entirely obsolete from the perspective of the market. Like, all jobs are taken by AIs, and all values produced by AIs, which drives the the value of human labor to zero and and, that, you know, as a consequence, drives the the the value of capital sky high. And so there will be a few actors who controls the whole system, and most peoples are entirely sort of dependent on them. And in some kind of cases, slavery, whether it just sort of kept alive by some, you know, substance stipend or, UBI, thing. And I would also imagine that in this economy, all the values of physical material goes very, like, it goes up quite a lot, as it relates to engine capital. And so everything digital drops in value. So you it sort of looks a lot like ready player one, right, where you have people just looking at the airport all day, living in some kind of slum, basically. And any kind of real meaningful agency, is sort of eroded. And and, it's a very drab existence drab meaninglessness existence.
Yeah. I I can imagine. So I mean, the time we are leaving, there are also some, I will call them cultural problems. Like, there is this digital divide. I mean, a lot of people are working I mean, some people are working on very specific and and innovative solution while the rest of the people I mean, I know people that, maybe they tried to to be t for the first time, like, I don't know, last week.
Yeah. Yeah. No. There's definitely that also. You know, the people who are able to understand how to work with these systems and the people who fall behind, And that cleft will just be massive, where you it's almost like you'll have a a society where, like, the vast majority is just passive consumers. And then there's, like, a few people who understand how to work with these systems.
And, is there any project, I don't know, on on Internet, that it, you thought was very interesting, and in some way, maybe it was, let's say, aligning with your project? Or
worryingly few, to be honest. I mean, I think there's there's a kind of a growing awareness of the same kind of issues that we see. For instance, there was this post called gradual disempowerment that came out a few weeks ago, that got quite popular. And now there's another one called intelligence curse. And these are by, like, AI people who who worked at Anthropic or other places. And and so there is a kind of a growing awareness of the issue, which sort of maps to what we think is is happening. In terms of, like, the solution space, I think it's a little, yeah, it's just there's not so many other projects that we look at being like this that definitely closely aligns with us. I mean, there's good work being done by researchers here and there, but I there's no coordinated effort that, like, very closely aligns to to what we want to do. I mean, there are there are things which are in the same ballpark. So there's, for instance, the collective institution the collective intelligence project, which are trying to also reinvent this kind of democracies with AI. And there is radical exchange, which to some extent are trying to do something similar with with markets. But, yeah, other than that, I don't really know.
And you mentioned some, some paper that I didn't know. Is there any other book thinker scholar that inspired you?
And Yeah. I mean, there's there's a whole series of, philosophers that would build on. And I think the main one that people usually don't know about is a guy called Charles Taylor. He is a, a philosopher from the seventies so kind of early on critiquing this kind of rational individual utilitarian basis upon which a lot of modern societies build, but with a sort of a, a concrete alternative. And in he and in his case, he kinda talks about how certain choices have, are just kind of an expression of taste and certain choices, say something about how we want to live. And so our our way of distinguishing preferences against sources of meaning or values is very much inspired by his work.
Okay. I actually haven't read, any of his work. And, is there, like, any other, like, or this one was the main philosopher behind the core idea?
Yeah. I mean, if people want to go deep, then we have a a paper that explains the whole process. And, there's sort of a section on the background where there is there's several philosophers that it's inspired by. Charles Taylor being the main one. Another one is called Ruth Chang. But I think it's easier to just go to the paper and read that. And if you wanna go deeper, then you can follow those those leads.
Okay. And, I have a I have a question. So for more people, for a lot of people, when they start the project related to civic tech, something else, it's a struggle to raise money. And, I mean, you had an, I would say, like, an an important collaboration with with with one of the biggest company in the AI world. And, like, do you have any advice for people that are building something that they think that, what they are working on is really valid, and, it could be helpful for all the humanity.
Yeah. Well, there are there are a bunch of, first and foremost, like, places that like to find things like that. SFF is the the one that comes to mind first and foremost. There's LTFF and a few others. But I think maybe more importantly, a lot of projects in this space don't really think so much about theory of change. Like, how does change actually happen in the world? And I think we we do need to upgrade our thinking there collectively. We've thought a lot about this also. And, I think the current sort of working answer is that for this kind of lasting deep institutional change, you need to have some kind of coherence amongst the experts that these are the right things to do. You can't just have kind of individual ideas floating around. And those ideas need to be reified by, like, working prototypes that make it very clear how exactly this this functions or or some working demo or something like that. And, and only then you can really go to the public and sort of have them demand that things work like that. And often, I see project going too quickly to the public, like, with climate change where there was no clear exact sort of way to implement or to some kind of policy decision around what to do about it. And so the the the didn't work. And, and so I think that there's a lot of projects should think a little bit about exactly what what is the fear of change. And it could be something also just sort of more local, like, there's a concrete problem we want to solve in our vicinity or in our community. And if so, that's great. But if there's no kind of clear idea of, like, oh, it would be cool if everyone use this, but I don't really know how to get there. Then, then, yeah, I think that's that's that's why maybe a lot of people might be reluctant to throw money on something.
Is there have you worked, like, with institution, like, institution have tried the your platform, or right now is more for, let's say, technical people or researchers.
We we would love to have more institutions try it. We or we're a little spread thin as an organization, so we haven't ourselves been able to kind of lobby for it. There was a point where we considered, running a campaign in San Francisco and using this platform for homelessness to kind of surface what people thought was important with homelessness and at least surface a little bit with policy makers there. And, but it's just been, yeah, we don't really have the capacity for it. But if someone is interested in doing this with either an institution or a community or whatever, then we're very happy to support them because we do want to see more use of this thing, but we just don't really have the capacity in house.
Okay. So the reaction from policymakers was, that they're not understanding how the tool could,
be useful or applied. No. It's it's more that we didn't really have the bandwidth to to have a bunch of dialogue and and run that whole project. So we kind of cut it down and then did other things or prioritize other things. It's just a matter of resources in house.
And, about the Webtree space, that is also quite active, have you had any, I don't know, contact?
No. Not really. I don't think there's any obvious sort of, overlap with what we're doing in the Web three world.
Okay. No. Yeah. I'm thinking because, Web three is, of course, more blockchain while you are working on more on the AI side. But I'm thinking about this, coordination aspect and also, the Webtree is trying to find new ways for governance. And so I think that in some way, this could align.
It it might, but I think that this there's there's a kind of a okay. I'm gonna I think there's, like, a little bit of, like, two little search maybe in the Web three space. Like, people don't really search for, like, what what what actually is the problem, what actually is not working, and, like, what actually have been tried. And people are a little too excited to kind of use blockchain as this kind of hammer, to just smack things with. And so there's, you know, there are there's a whole field called sort of choice, right, where people have sort of thought about how to take, like, a decision for a long time. And most of the three people who are building coordination take us not really even aware of it. I'm not even saying that that field has necessarily the best solutions, but I think there's a kind of a lack of doing the background reading maybe to find, what actually is not working, like, what actually are the problems.
I absolutely agree that there are a lot of things that can be improved, also in the Webtree space. And, and so you were suggesting, like, more, research, deep research instead of just the
Well, at least trying to understand, what problem are you solving and why and what has been tried before. And if it's been tried before, why didn't it work? Something like that.
So I I'm thinking about history. Yeah. I like to study history for the same reason, so I know what,
what is As we should. Yeah. More of that.
Yeah. Yeah. And, I mean, I have another question that is, about the issue of any message, for the people that are building a new kind of solution that are exploring, new ways, possibilities, like you're doing?
Yeah. Maybe one thing is, like, take it seriously. Like, there's there's just, like, a a massive need for it. And, like, I feel like a lot of people are sort of half assing it a little bit. And it's it's it's unfortunate because it's a really important project like the you're doing important work. So, yeah, don't yourself and and and take yourself seriously maybe. It'd be a one way of framing it.
I I agree about the importance of, yeah, trying to find new solutions and experiment. I don't know if you have any any questions or, like, any other kind of thoughts to that you would like to share.
Nothing comes from my email.
Okay. Because, actually, like, to ask more question, I will have to actually really dig into the into the platform and also exploring the the repository that you said that is open source.
Yeah. I think the best place to start is probably to read the paper. The paper is called what are human values and how do we align AI to them. But even though that's the title, it has a lot of good stuff about coordination tech and civic tech, a little bit between the lines. And especially the sort of background sections and and the method sections, I think are, yeah, part of parts of the method sections are are interesting regardless if you're not into
Thank thank you a lot, for your My pleasure.
Yeah. Alright. Take care.
Thank you.
