H-AI-L to the Chief

Transcript

All right. Bonjour. Good morning. Sorry for the the mixup and the slight delay there. Uh, I'll I'll jump right in. Hey everybody, my name is Lane Reig. I do research at Near Foundation and I am really excited to be here and to talk to everybody about the intersection of governance and AI, two topics that are near and dear to my heart that I'm very excited to talk to you guys about. Um, for context, I uh joined the Ethereum community in 2017. I was formerly at the Ethereum Foundation as a core developer back in the day and I I helped build um a layer 1 project called Space Mesh uh before joining Near about eight months ago. And um the thing that originally got me excited about the

blockchain industry, about Ethereum in particular, was not so much the financial side of things. It was not so much the the tech even uh it was really kind of the human side of the story, right? I I like to call blockchain, Ethereum an operating system for building better human institutions, right? Institutions that are more transparent, more participatory, fairer, more open. Um, and that's uh what I'm working on today in a project called House of Stake, which is a novel full stack governance initiative, a DAO that we're building to govern the Near Protocol. Um, and we're leaning very heavily into AI as well. So, I'm excited to share the work that we're doing there.

I want to start with a visualization exercise. Close your eyes. Okay, you don't actually have to close your eyes. If I ask you to visualize powerful AI, right? Human level capabilities, what do you visualize? What do you think of first? If you think of something like this, you're probably not alone, right? This is to be expected the stage we're at right now in the AI journey. Right? AI is still new. It's mysterious. we're not sort of fully uh aware of its capabilities yet, right? So I think it's kind of natural to feel that way. And if I ask you to add governance to the mix, okay, so what is very powerful AI with a role in governance look like?

I couldn't help but put up this image. This is obviously a very old AI generated image from a blog post I wrote about this topic a few years ago. Um, but if you kind of imagine a world where AI has a central role in governance, you know, where humans are answerable to the quote unquote robot court, the AI court, I think that's still pretty dystopian. Um, but I'm here today on the stage to argue that actually AI has a very constructive role to play in governance. Doesn't have to be so dystopian. Some of you may have seen this article in the New York Times just a few days ago. And you know, in typical New York Times fashion, it paints a pretty dystopian picture, a pretty bleak picture of the role that AI is playing in governance

around the world today. Um, namely that it's beginning to erode democracy in their words. Hope the videos work. Right. So they actually site examples as you can see from dozens of countries all around the world about deep fakes, AI generated images and the disruptive role that AI is beginning to play in in in democracy and in governance. It's really quite a frightening situation. In particular, deep fake videos like this one are increasingly common and platforms like Tik Tok X uh Facebook, etc. They're pulling thousands upon thousands upon thousands of these images down and videos, right?

that that you know their their policy officially states that anything that's AI generated, AI edited has to be labeled as such. But of course these things are not. Um so this is a very famous widely circulated example from you know the most recent United States election. This is really just the beginning right of of this kind of disruption we're going to see in democracy from AI. And in a very famous and very particularly extreme case actually an election that just happened a presidential election in a major European country in Romania in December was actually cancelled. I don't know how many folks here are aware of that, right? Because of severe disruption from AI into uh the integrity of the election. So this is the the the Supreme Court said that there was a quote impact

impact of extensive deployment of artificial intelligence among other things. Interference by Russia um a a unknown candidate who appeared out of nowhere and suddenly captured a large portion of the vote. Uh and again they literally unnoldled the result of the election and they haven't even held the election yet. So this is not hypothetical, right? These threats are real and present. They're happening today. Okay, so that's that's the intro. Things are pretty scary. The situation feels kind of dystopian, but again, the reason I'm here is to pose the question, does it have to be this way? Right? Uh can't we do better? Can't we find a not a destructive but a constructive role for AI in governance and democracy?

So, let's look at some, you know, positive use cases, some constructive uses. Uh some of you may be familiar with this gentleman, Francis Suarez. He is the mayor of Miami and he ran for president very briefly last year in the United States. And he actually created an AI digital clone of himself that his constituents, his voters could interact with. They could ask questions and it would, you know, share his policy, that kind of thing. What I like so much about this example, number one, is because again, it's a it's a constructive example of a candidate leaning into AI and using it to their advantage. And number two, it it's constructive from a Democratic perspective, right? Because it it gives his constituents more access to him. Um the number of candidate,

sorry, the number of voters that a candidate can speak to in a given day, it's obviously very limited, but an AI clone can speak to a theoretically unlimited um you know, number of voters. And uh I know that looks a little bit low res. It's 2D. The mouth movements are a little bit unnatural, but I wanted to share this clip with you. So, I was at an AI conference a few days ago and uh this is a photorealistic 3D lifelike digital clone of a K. This blows my mind. Uh these are coming, right? So, this company, the company that created this, it's not just a visual likeness. It's also um they do what's called a brain clone as well. So, actually they train it on speech patterns, on you know, things someone has said or written. They're working right now with professors, with leaders

of companies. Um, and I think we're really like one election cycle away from you being able to interact with a lifelike photorealistic 3D clone of your favorite politician, which is either exciting or terrifying depending how you look at it. But again, I think there's a very constructive role here for AI to play. So moving into the kind of onchain world now to share a little bit about what we are working on um at Near and at House of Stake. We we already have actually built a AI co-pilot. So shout out to the MetPool team. Thank you for their help with this. We're really really excited about this. Right. This plugs directly into our governance um interface. You can actually interrogate proposals. You

can ask it what proposals are on the table. You can ask it uh how you should vote, what it recommends you should do. Um you know, you can ask about kind of the upsides and downsides of any given proposal. You can ask it for help writing a proposal, reviewing a proposal. Um and you know, I like to kind of think of there being an agent agency slider. Okay, so this is like close to zero agency. This this is this co-pilot, this this quote unquote AI agent is not actually doing anything on your behalf yet other than helping you. Um, but so it can't vote for you, but you can tell it how you want to vote and then it will sort of pop up your wallet and let you take that step. Okay. So, this is kind of like stage one um of of the AI governance tools that we're

working on and we're planning on and indeed already working on expanding this into a whole suite of AI governance tools with some amazing partners. This slide, the next few slides come from our friends at DeepGV. So, Deepgv is a project that uh has also been pushing the envelope in using AI in sensemaking um on both onchain and off-chain use cases. Uh some folks here may be familiar with Audrey Tang. If you're not, I really highly encourage you to uh look her up, listen to some of her podcasts and and talks and and read some of the books and things she's written. Um she has this amazing vision for this thing uh called broad listening, which uh allows us to escape from the information explosion that we're all in the middle of right now. Right? Right?

So that leaders, candidates, um, anyone in a position of, you know, decision-m or authority can actually use AI tools to filter, to cluster, to make sense basically of sentiment in a community. And this is also something that that they're working towards and that we're working on at House of Stake with Deep Gov. They uh put this in practice in the most recent Gitcoin grants round. I'm sure many of you are familiar with Gitcoin grants. actually $50,000 of the prize pool in the most recent Gitcoin grants round was um was allocated using three AI politicians. Right? So again, now slightly more more agency here. They uh did a kind of sentiment analysis, a clustering exercise where they created these three different policy documents

to correspond to kind of different political factions within the Gitcoin grants community. So you have one that's a bit more right leaning, one that's a bit more leftleaning, one that's a bit more centrist. And again, they they actually uh actively allocated um public goods. And we're planning to again do this very very soon in in House of Stake. And they went a step further and the Deep Gov team actually just came back from an in-person event from a governance hackathon in Bhutan uh where they're they're using the same ideas, right? So they're doing this broad listening technique using things like a telegram bot where any citizen can authenticate using the national uh identification the national ID um and then share their thoughts ask questions etc. right and the output here which you can see on the right side is a really

beautiful um kind of structured analysis of the preferences the clustered kind of beliefs and preferences of different groups of people the strength of those preferences breaking down very kind of granular views. This is an extraordinarily powerful idea and we're really really excited. I think we may be the first DAO to to to build something like this. Again, shout out to Deep Gov and thank you for their help with this. This is my friend Abraham Lincoln. Um, you know, the next step after we've done some of this sense making and deep listening is to give our AI tools more agency, right? Moving along the journey along kind of the agency spectrum. And uh the vision here is that these AI agents will have a seat at the table

metaphorically speaking in governance at House of Stake. They will actually be able to receive delegations from humans and vote on their behalf. Right? And we're we're kind of a few months away I think from from executing on this vision. Very very exciting. They can introduce proposals, vote on proposals, etc. There's a deeper why, right? So it's it's worth reflecting. You know, all this stuff sounds very cool, very exciting, right? But like why are we actually here? Three reasons. Number one, governance is just is just critical. Governance is so important, right? We technology is not enough. At the end of the day, it's kind of people, human systems all the way down. And in order to build things that scale, build things that last. We need robust governance systems.

Proposition number two is that humans are just bad at governance. Sorry. You know, like we have been dealing with the same issues for thousands of years. You can read Aristotle and Plato and the things they talk about sound extremely irrelevant today. We are short-sighted. We're narrow-minded. you know, we are selfish. Uh we don't always have background information or context. We're not good at explaining why we make decisions. And the proposition number three, which is maybe the most important one here, is that AI might be better at governance than humans are. Um we don't know for sure, but it's an experiment we need to run. Right? The AI is always on.

Uh it always has context. It always has information. Um it's much better at processing large amounts of information than humans are. Right? So this is the experiment that we're working on right now. Um I'll just quickly move through you know the the flip side here just to present the whole picture. There are some risks right and these are not specific to blockchain or crypto. So the first one is alignment. Um even to the extent that AI appears to be acting on our behalf and understand our interests the reality is that uh it's not actually doing that. It's not reasoning and it's not doing it from a perspective of moral or ethics ethical code or values or something like that. So there's risks here. Um bias right? So humans are obviously very biased. We know AI models are biased. Uh they see the world

differently than we do. But I do believe that they may be less biased than humans are. And explanability, right? AI is not very good today at explaining why it makes decisions. And I think this is especially critical in governance. Um in order for the human stakeholders to feel that there's legitimacy in a governance process, they need to understand not just the outcome, but how those decisions were made. Right? So these are all areas of active research. We're working on identity. I'm short on time so I'm going to kind of flip through these very quickly. Uh democracy on chain is very hard as everybody knows and there's projects out there like world ID that are working on this but we are actively doing research where we believe that we can actually build um kind of unique human systems and potentially even democracy on chain

using AI agents uh to act as guards there. Um you know the the vision here is to build an AI CEO. This is actually Ilia's vision. Uh Ilia wants to retire and uh you know the vision here is to kind of remove the burden on the human agents. um in the governance process and put most of it, not 100% but most of it on the AI instead. And my personal vision for all of this and this is probably the most important idea here is not a single AI agent, a single AI CEO, but it's actually a swarm intelligence. Okay, so the vision here is that every one of us onchain, offchain, in any governance situation, we have an agent that is aligned, that understands our our best interests, our our desires, our values, and is just out in the world acting on our behalf. So we can sleep

well at night knowing that it's out there voting for us. Right? This is very explicitly the end state, the goal in my personal vision of what we're trying to build at House of Stake. Um I didn't get much chance to talk about House of Stake. I've thrown the name out there a bunch of times. We're building a DAO to govern the protocol. Goals are improving economics and network security. Uh you know, quick shout out to everyone who's been a part of this project. We're very grateful for your participation. We held an in-person event here two days ago. uh which is huge because I think you need to get people in a room together to build trust even in a DAO. So that's how we're going about doing this. Uh we're doing a call for data governance data to build the world's largest data set uh governance

data set to power all of the AI infrastructure and applications I spoke about a moment ago. So if that's something you're interested in contributing to, we're working on this with a research lab at a major research university. Um we have some complicated research proposals. So if that's interesting, join us. This is a great time to get involved. Um, my DMs are open. You can find me easily on Telegram or on Twitter. Um, we have a governance forum as well. Um, yeah, this is a great time to get involved. So, please reach out. Thanks for your time.