So we will collect recordings of conversations of a few predetermined questions. How can we use resources to make your community better, etc. The summaries will be generated, say, the next day where they will gather. They will see the summary and every person can give feedback on whether it’s ...

And it has to stand for itself because no one else will look at it. So knowing that that is the kind of person I’d like to empower. So what would we give her? I’m trying to steal in a conversation with her. She would say, OK, I have all ...

Technically, no, she’s not a technical person. So it has to be like just “go to this website.” So I’m thinking of her just as an example. She could really rally a lot of Detroit. I’m particularly interested in cities like Detroit that are, in America, it’s like purple cities. ...

I mean, Shagana, for example, who runs the Heal Machine group, she gathered like 200 people. She put us in touch with just 15 because we want to start somewhere. It would be very easy for her to be able to do something like this. Easy as in socially.

Another one is to create this in a way where this directly feeds constitutional AI.

Sure. I’m thinking about what the suggested pipeline should be. For example, it’s one thing to tell people, can you rate your summary on the accuracy and feedback loop?

So I’m curious, what would be the ideal stack here? Or what would be the ideal approach? Because you were saying you’re trying the constitutional approach. Our relationship is a way to do this. It’s not bad. There is thinking about, you know, directly having graphs and graph representation speed ...

Facilitated discussion because you will find similar opinions in their different pockets.

I would like to hear your advice on which models do you think, because this is the kind of experiment I would love to run. Honestly, I would love for there to be a couple months of my time next year where I also try this in different communities. I ...

Yeah, much smaller pockets would be good. How would you engage the population? Would this be a paid engagement? Is there an incentive?

Interesting. How would you anticipate and I guess I start from Taiwan. I’m thinking much or even smaller than Taiwan I think.

I’m not at all against working. I don’t know who’s still around in that group. I will reach out to a couple of my friends. I talk about them in my thesis. I’m not against working with them. I just know that they are not very friendly themselves. But if ...

What are your thoughts about. During COVID, they shifted to a lot more digital stuff. And like everything has to be in person. Up until then, they have been quite against collecting smaller snippets from people. If you can’t actually get all of them to talk to each other.

So I have this. When I was writing my thesis, I talked a lot with Jim. One of the things I was saying is I’m really interested in AI tooling. And this was back in 2015, 2014. And he was not very open. I’m not looking for your blessings anymore. ...

No, no. I think there’s a. Yeah. They were very much the first movers in the landscape. Jim Tishkin himself was charted out a lot of deliberative approaches. And I think there has been some negative side effect of that to prevent other approaches to pop up. By creating a ...

Yeah, I have a lot of thoughts. Oh, a lot of thoughts.

I haven’t seen the URL at Stanford. These are my people. All of their websites are the same. But yeah. Jim Tishkin’s group is this. Deliberated democracy. Yeah, this is Tishkin.

I’m curious. I haven’t talked to this group much, actually.

I just I didn’t know if you knew or not.

This is like where I was. I think that’s interesting. No, no, no.

Any other thoughts that come to your mind?

I need to learn traditional. I have studied Mandarin a little bit, but I don’t recognize traditional characters.

I mean, I’m actually quite interested in becoming also Taiwanese.

How do we make more Taiwan’s around the world?

Yeah, I think that meaningfulness is very important. And that can only happen with smaller scale contextualization. I’d love to share this with Colleen. You’re still recording, right?

And I think in fact, being able to say I don’t actually have a lot to contribute on this question is a big part of civic responsibility. So if anything, I think the way democracy has evolved in the last two centuries created this illusion of everyone has equal parts ...

And I think it’s also that, I don’t know, if you ask me for biosecurity and AI, what’s your action space? I have no idea. I’m not an expert. But I can think about this. If you ask me my outcome preferences, I would probably have pretty robust answers. It’s ...

I like that a lot.

I’m curious what else could we do that makes this contained in one process, in one product. I don’t know what I’m looking for here. That’s why I’m at a loss of words. I think this is the thing we’re trying to make headway.

I’m thinking about it from a build perspective. If we were thinking about how can we have… Because tools like Talk to the City or Models are not trying to solve the how to scale coordination. It’s a tool that does one thing. I’m interested in how can we have ...

If we were just doing a hackathon to make a prototype that’s just one, two, three, how would you approach it?

Ask people what’s future? What are your thoughts? One step away from that kind of question. Or what should be top of the agenda? About what actions are possible? I’m interested in using the Asian inference for thinking about helping people think about what could happen. But you can also ...

Right. So I see Talk to the City as just one of the many experiments towards this. To me this is the North Star. Talk to the City is fun because you can use it for three outcome preferences really well.

So you’re saying civic tech should have the head start with better tech? And at that point the government is just responsible…

How do we not just throw people referendums but actually a flow through which they can suggest actions, suggest, you know, I think this is what’s likely to happen with an action and suggest outcomes. And my goal isn’t that like all of this needs to happen in Silico. I ...

Like this part I am particularly interested in. Like I don’t know my preferences. Like asking me to have a preference over a possibility distribution that I’m not familiar with is almost a disservice. So, to me, the important part is given people don’t come with strict preferences over outcomes ...

That makes sense. That makes sense. I was exploring this for Talkative City and then it hit me. We don’t need this. We can just work with CSVs for now. I don’t see Talkative City to be used in the immediate future about an issue that requires contribution with anonymity. ...

Great. This is good.

Right. You have to show it. Like, there’s already something here.

It’s not necessary.

Anything promising that you have seen?

Have you listened to a lot of the ZK work on this front?

So, this is where Europe is going right now.

Yeah.

Yeah.

Is there something specific that actually would be helpful or we will have more data than a human would be able to read in these systems. So what would be helpful?

But people be canvassing would work in your projection. Like, what is what is something that we can do that actually would bring value today is a very like I like to use that compass to be able to go to instead of talking about.

I would be interested in your thoughts on how AI advanced.

OK, that’s good. I’m curious, like what comes to your mind is, you know. Which features would be most interesting or valuable to look at next.

Oh, great.