Building Collective Intelligence Tools for Human Alignment with Nicholas Brigham

Transcript

foreign [Music] ation on the pods today we got Brigham Adams who is the founder and chief scientist at goodly Labs which is the r d lab focusing on developing collective intelligence systems for a self-improving society that work includes training AI systems and building pro-social software to provide social structural scaffolding so that could improve Society so I met Brigham at the plurality conference in San Francisco a few months ago and I just found him to be a really talented facilitator of the conversations that we were having there but also um I just really respect people who have the ability to make plays and to be an

entrepreneur and to ship software but also the theory the sort of synthesis of all the things that are happening in the world the current trends and um basically having strong a priori knowledge theoretical knowledge and then empirical knowledge that's derived from experience and so fun episode bergam is just a really Galaxy brain we talked on this episode a lot about lots of intelligence we talked about how AI alignment requires human alignment we talked about pro-social software versus anti-social software and I think where it connects to green pill is how a web 3 world outfitted with pro-social technology could help create a more regenerative economics movement so really fun episode with Brigham and I think that you're going to enjoy this

episode Valley Dao is an open Global Collective working to finance and democratize the governance of synthetic biology Technologies to protect the future of our planet research and progress into synthetic biology is critically constrained by the world of nation-state regulations and bureaucracy from middlemen but the potential that synthetic biology has for the world is massive it can produce a world in total harmony with nature we can grow stuff using cells like software it's awesome and the best part is we no longer have to pull carbon out of the ground to grow and maintain civilization don't you think we should be funding this Valley Dow does and that's what they're here to do to try and make the solar Punk World of Tomorrow a reality today so so join Valley Dao right now and save the planet

at valleydow.bio hey Brigham how you doing I'm doing pretty good today it's a it's a beautiful day in San Francisco actually let's I'm wrong I'm not in San Francisco I'm in Topanga Canyon wow I'm already I'm blowing it here though if you're if you're extremely online then you're connected into the metaverse and that's where you are at least that's how I feel all the time that's right I'm I am ever right now yeah okay yeah uh well I would say to talk about goodly Labs you and I met at the plurality conference a couple months ago and was just impressed by the work that you're doing so could you take us down the the rabbit hole what are you working on yeah um with goodly Labs uh it I rarely saw it in the world we needed some way to

get the people who kind of know about social social reality through social science working together with the people who actually build stuff at scale that can touch everybody in society um so yeah when I was getting my PhD at UC Berkeley I just you know I noticed this guy Zuckerberg is like holy crap he's got a billion people dancing as dance and in just under a decade to take a thousand years um so and and you know there was nothing particularly um wonderful or magnanimous uh or like collectively intelligent about Facebook but if we were actually designing that stuff to make Society better to help people coordinate cooperate at scale wow

like what what kind of world could we live in so um yeah I I put together goodly Labs because I saw that no one else was really building the collective intelligence infrastructure that could allow us to be like reflexive and sociological and really smart as a society uh so yeah the header on your website says youth building human capacity for human alignment and I love that like every once in a while you meet someone who says back to you what you think you're working on but in their own words and that's how I felt when I was uh on the website well I've been feeling the same way like seeing your tweets and stuff lately and it you know it feels good to know you're not like alone on the battlefield yeah um I actually kind of switched up the tagline recently

because we realized you know this is the big challenge of our time um so many people are talking about AI safety and AI alignment um but alignment to what you know how do we even Define what it should be aligning to and right now the systems we have for just basic human alignment um are are failing they you got these political policy making systems that sort of made sense in the 18th century um back our best Communications technology was like pigeon or a horse and that's what we're using we're using and you know everyone needs to come to one town in a swamp and talk one person

at a time in a giant Hall where 435 people that is that is really old archaic technology it's really anachronistic um in a world where yeah we can we can bring Global Supply chains to your doorstep and we can have each person could be theoretically talking to a billion people the many diminished communication 24 7 365 asynchronous communication we could be doing coordinate and cooperate that there's just such a big dealt between what we could be doing and what we're doing um but yeah we just want to build that capacity so that we can start aligning so that we can tell to AI you know be like this this is what makes humans

happy I got I got two takes uh that I'll take in the and put in the pot and we can stir it from there the first is uh this idea from Edward Wilson that we have Paleolithic emotions medieval institutions and Godlike technology this felt like wow in 10 words such a concise way of uh talking about modern modern governance but the other one is um I take that I saw uh from Daniel schrockenberger who's this meta crisis alignment person talking about how basically everyone's talking about artificial general intelligence alignment and how are we going to align the AIS with each other and we've already got a decentralized alignment problem which is the seven billion humans on Earth we've already got a GI intelligence problem a general

intelligence alignment problem uh would be existing Humanities and now you're just taking uh agents that are a thousand times smarter and for now are embodied and putting them in the mix with each other so um you know I like this idea you were telling me about before we were ordering about AI alignment uh meaning solving human alignment first or the way those two ideas interact with each other uh maybe we could pull that thread a little bit yeah absolutely I can you know I can start with a concrete example although we can talk about like more more ways to address this big issue so there is clearly a human alignment problem right now and I and I know smack the burger talks about this um and and the social Mama talks about this there's the human

alignment problem of um first political parties and then social media companies heading from the tractoring of reality um if if if a political party can feed you a certain slice of reality um it can get you to vote for it and if a social media can feed you an even narrower slice it can keep your attention and and keep you kind of like these debates um so this is about how we form like common knowledge how do we all share the same understanding of reality yeah because the the 20th the 20th century institutions plus you know if you want to throw in social media as a new 21st

century set of Institutions they're all kind of sorry I didn't mean to interrupt you I was gonna say I'm not ready to cope with calling social media institution yet but uh I'm gonna have to digest that one as as you're speaking okay for interrupting you continue okay that's okay I mean but these these institutions are kind of feeding these dynamics of narrative Warfare um where it's more important for people to to gesticulate and show what position they're in vis-a-vis other groups it's more important to do that kind of social proof um stuff than to actually get down to the heart of matters get down to the truth um and try to understand you know the the consilience the multiple different

perspectives on the same um object I'm an eye in the world so one of the things that we're working on to try to build consilience to build shared meaning to build what we we call shared reality um is just to give people the tools to easily stream the ways that we manipulated them and I don't even like really hardcore propaganda disinformation manipulation it could even it could even be like a gentle way that we fool ourselves there's all kinds of biases that we shouldn't fall break to um and so you know we're looking at things like appeals to Authority we're looking at you know once the one wrong argumentative fallacies ignoring

selection effects um you know fallacy all these ways that we do men School ourselves we can label those directly onto the words and phrases of the content that we're uh consuming on the internet and when all this content now you know instead of just reading a news article and letting it all wash over you um we now can interject the appropriate doubt on the particular words and phrases that are committing these different errors so that when someone's thinking through you know it's not even that they're reading an article and and something at the top of the article says this is fake news disregard this entire thing

um which is like too much into the narrative Wars but with these like very uh potent and specific introduction of doubt about a particular concept really is and just realize like oh I kind of agree with this person but they're also schooling me um this person's on hide but they're actually using the trick um or the person's on the other side and they're definitely using a trick if we can get more people to realize what those tricks are how how we're getting fooled how we're getting manipulated how we're manipulating ourselves in each other um we think we can start to constrain people into uh at least a way of sharing

and talking about more and more of what we what we could call reality what are the inner subjective reality if we're not moving each other we'll start to get a better handle on what that is and then we can we can at least achieve like um a high fidelity disagreement and know what we're really doing about what we need to compromise about I love that we're talking about High Fidelity disagreements instead of just arguments or what are they called Flame Wars online they're two very different ways of framing people having Different World Views so we would just want to make it um possible for to live in a world where people are interested in this more High

Fidelity communication um and so we've built tools for that and then we just want to we want to spread it as far and wide as we can so um all of this because manipulation is tricky all of this actually requires humans doing um cognitive tasks like micro tasks like mechanical turkers like like online crowd work although um we use some really sophisticated interfaces that we've created um they do these tasks and once the humans have found you know like 10 000 examples of the false dilemma or 10 000 exam during selection effects we can then actually feed that into

um machine learning systems that create AI that can also label those sorts of mistakes so it starts with a human process but then the humans end up training the AI to do something very sophisticated that they can't currently do um so it actually is a super intelligence that humans can develop that's better than AI super intelligence but the kicker the really awesome part of this is that then we have an AI that can do this sophisticated thing but we also we also know exactly how the AI was trained we understand it we can tune it we can revise it we can improve it we can continually improve it um so we're not like the one of the

things I'm worried about talking with folks in the Bay Area there are some people who are really bullish on AI well no it's going to be great we'll we'll just hand our problems with AI and it'll solve our problems for us um I want our problem solved as much as anyone but but I I kind of want humans to maintain responsibility for ourselves like I'm I'm not capable with the world where um we're just kind of like the pets of AI remember we hope that it remains nice to us um it also just submits like what what's the point of a human life I mean what's so cool about humans is our agency our ability to

use the world and our lives and our surroundings and if we're just handing that off to the AI like what are we here for yeah I I was reading the uh the The Singularity series which is this book of uh it's a four four book series uh that takes place in 2025 2035 2045 and 2055 over the course of the series you can see how uh AI takes over society and one of the sort of like plot lines that I thought was interesting was um humans wanting to retain their agency but um the AI like the AGI works out that if they uh the humans are gonna blow

themselves up and so the AI uses that as a pretense to like take over all society and keep the humans from destroying each other um and I thought it was just kind of a a interesting of course science fiction not predictionary in any way but a way of thinking about like you humans kind of even don't don't deserve you or don't know what to do with your agency was kind of like the main message I got out of that but um I think it's a I mean critique I mean we haven't really proven ourselves have we I mean and we and we're we've got these amazing technological building blocks just sitting here you know we could be we could be governing ourselves in really amazing direct Democratic ways

um where we're you know 24 7 365 we're deliberating about the important issues uh how how do we coordinate how do we um solve problems together and delegate our resources and we could be doing that we could be determining online right now with existing technology we could be determining like a long rank or list of our of our needs and our wants um we could be coming together and and we are a bit and I'm glad that you're helping lead this too you know we could be coming together in daos and big distributed networked organizations and actually structuring the functioning of those in new ways that are that are like

highly effective and not so hierarchical we could be developing the tools for human alignment and then using those tools to prove that we can come into alignment like all of that we we keep doing right now and we're kind of there's like a few people in the world who are kind of on this and a lot of them are probably listening right today so uh thank you thank you everyone who's already on this let's with our people right I'd hope so and uh this is my chill for if you know someone who should be listening then don't forget to like And subscribe now that I'm a podcast host I have to always say that share this with your friends if you think that they uh if they can help out here but um I'm yourself saying it just at the pub

and you're like oh wait I know yeah no um the the thing that has bled out of the podcast into my social life is that I started calling the audience coordination uh because it was like this is a sub podcast so thankless and Ryan and David called Their audience bankless nation and so I started calling uh everyone coordination and and now it's kind of become a meme for like a regen Network State uh out at the pub so um a coordination is a regen Network State and that's the one that's that's bled out of the podcast so um yeah but uh um yeah building movement is hard work and it takes persistence and you know one of the things that I was really excited about with having you on Brigham is that you know you kind of have like a really interesting philosophy and sense

making capacity about this but you're also making it happen at goodly and I'd love to spend most of the rest of the episode just going back and forth between the theoretical and the empirical um so can we take a dive into what goodly is actually doing to to stimulate this movement and what tools you all are are putting out there and what kind of results you've seen yeah totally um so there are kind of three projects right now that are scale ready and I can I can talk briefly about any of those and pull on those threads and go deeper and what they all have in common is um that they're all in one way or another collective intelligence tools and the way it redefines is is you know it's kind of cute it's just it's just

like any any tool that multiple people can use um the where the output of it is like you get too heavy better than one or two thousand heads are better than one two million heads are better that that is not a foregone conclusion you know like if you've been in a bad relationship your head can be much worse than what and there are all kinds of systems that we build and in western civilization particularly built lots of systems where it might be um there might be a corporation with 50 000 employees but that the thing is organized so that the one person that top can control everyone um so they're actually they're they're

created in such a form to make sure that you don't have a situation where two thousand heads are better than one or more powerful than one in that system um but that's that's like the basic idea of collective intelligence um and it's not easy to accomplish there are lots of ways to screw it up uh but what we're doing yeah we're having success with public editor already um this is the the manipulation labeler where we we feed news then and then what comes out are the articles with labels on them and scores from zero to 100 and King The credibility of the whole article but also the score for each label indicating each mistake in the article and we put

this system to the test um with the EU the EU commission has their own kind of uh disinformation experts and um they gave us a set of 250 articles that were Russian propaganda articles and they had already kind of identified the disinformation in them so we took those articles and we put them into our system um with a naive set of users of annotators you know these these people who are doing little micro tasks to analyze the article um and they're each doing a different task you know they're looking at the reasoning or they're looking at the way the evidence is used or um the way that probability is discussed uh or how the quoted sources are being called upon to to support arguments the

relevance of different arguments they're each kind of looking at these different aspects of the article but when you take all of their work and you Stitch it together you you end up with like an extremely intensive assessment and review of the article so we we took those 250 articles from the eu's team and we put them through our system with a bunch of naive users who had barely been trained and then we gave the Articles back to the EU team and said what do you think um and they they reviewed and said wow like in a hundred percent of these articles you you're providing really valuable helpful information for the reader and 94 of the Articles we found the very same disinformation that they

had found and in 86 percent of levels we found additional misinformation that they had not found so uh the stuff really works the next big question is like can we get 2 000 people doing 15 minutes a day of past and you know ideally that's volunteers and these are these are students and these are retired folks and they're stay-at-home parents all of whom have some investment in making sure that we can actually share reality and that we're there's some kind of curb on the extremism and the impulses towards extremism in the art of water

um so if we can get 2 000 people doing 15 minutes a day then we're covering like top 100 news articles look at this and those top 100 euros are going to become all the radio and television stories and then all the social conversations over the next two weeks so um we feel really confident that we can have a great impact um and our next problem and uh just rolling it out in Europe we're gonna start with English speaking Europe uh but we're already we're already getting pushed back I guess or maybe not pushed back it's more like well in multiple countries that are like no you need to

do it in France you need to do it in Germany you need to do it in Poland you need you know Romania really so um that's a system that we want to provide dude you know 15 or 100 different language communities around the world over the next five years where they each kind of have their own capacity to um rate and and assess the quality of the News That's flowing in their language across the web and then as I said when we have many examples of each type of mistake um we can do that for we can have an AI That's doing that for all content on the internet and we can in fact like sell to the AI companies this sort of governor

um with with this uh like strategy will no longer create or reasoning errors it will no longer manipulate be manipulative and the way it responds um and I think that's going to be you know that's important very human to alignment um it's it's going they slow down create kind of a forces towards your trip you need of the variants um uh and then you know we have some other projects that are like we allow do you mind if I can't find the disinformation project real quick yeah I had a question about it oh yeah so I guess I'm wondering um uh one of the

ideas that I've sort of discussed with other people who are working on disinformation and misinformation is like you know say you build a system that is is like labeling the disinformation and people are starting to trust it it kind of like if you if you're deployed like Zuckerberg style then it creates a centralization risk like an authoritarian person who can almost control what people think right and so is there any way to solve for that with decentralization or decentralized governance or like some sort of check and balance if that system gets captured by one political party or something like that because I just the audience here is is very decentralization heavy and decentralized governance heavy and I just I just feel like a lot of them are probably asking that question listening to that I'm glad you asked yeah the um

this whole system we we've been very careful as we've developed it and we'll be careful as we roll it out to make sure that it's totally transparent and it's and it's really neutral so um you know we've been testing it with uh students and uh people from civil society orgs and um just people off the internet I'm a Mechanical Turk and the way that we construct the actual tasks we have people kind of walking through they're looking at one smaller block of text in the article and they're walking through question by question like it's it's very step wise

um and therefore kind of like extracts the user from their biases they might have about the entire article so if they were reading the entire article they might realize like from the from the beginning like oh this is a a very um far left uh author and I consider myself center right so I'm going to trash this article right we don't we want to avoid that sort of be in either direction of course um and so we've we kind of break the tasks out so that they're pretty deep biased also each task is done by multiple people independently and then we're looking the algorithm that's

looking for a consensus among them and that that algorithm is looking for a consensus on each question and answer um but it also we have an algorithm that looks for consensus in the aggregate to see like are people consistently seeing this the way that other people from the other side of the aisle are seeing it yeah and we can make sure that um we're not that we're kind of like D waiting or adding D waiting anyone who's really biased in the way that they're looking at stuff or we can give them remediary of tasks uh the mechanism oh it's gonna reject the mechanism designers I hang out with we call that

uh cooperation across social distance so if two people on the right agree about something and two people on the left agree about something that's not as strong of a data point as someone on the right and the left agreeing about something that social distance matters a lot absolutely yeah so we yeah we have some passive measures for to learn like what someone's positioning is in the political ideological um and then uh we kind of distribute the tasks in a way that's trying to find uh agreement across those divides um and then you know in terms of recruiting people like we you just have to we have to place ads in multiple places right like we need we need ads on

break we need ads on The Drudge Report in addition to ads on on stuff that's more on the left side of the aisle so um I just want to make sure that we that we closed that Loop before we finish talking about the first product but I also want to respect that we only have 20 minutes left and you have two more products that you wanna that you want to tell us about so can we get into products number two and three yeah the the stuff as far as the stuff that's ready to go right now we have a project called demo watch um and this one it looks at this actually came out of my dissertation work at UC Berkeley so um I was studying social movements and I was studying like peace and conflict and the Occupy Movement happened and

um it was just kind of amazing looking at the literature and political science and sociology there were so many one-off case studies about like this is how things went down in this particular City with this particular movement at this particular time um but the there wasn't really great research that looks at a bunch of movements and understands and has kind of the comparative leverage and the rich enough data to understand the Dynamics of interaction between policing protesters so there was a there's a huge opening in that research on questions that take the form of like why did the police act this way this time or why did the protesters do this in this situation

um really just like not settled knowledge on that at all so um I decided I wanted to look at the Occupy Movement because all these movements were doing similar stuff at the same time uh for the the same sort of reasons and it sets up like a nice natural experimental design that's really good comparative data to understand why police department the city of Norman responded differently but it's fundamentally wouldn't happen to another question and in order to do that I I had to become a data scientist um because I was stupidly ambitious I my advisors were like yeah pick four cities you know pick six cities and I was like no no I want to study all 184 cities and

towns across the U.S uh so my hack was to teach the methods course at UC Berkeley and I I pulled the top murderous social science students and the social science we went and scraped all the news stories and all the review and television stories about all the Occupy movements across land and then we started analyzing the content and the tools for analyzing the content like just were not up to the task of doing really rigorous and intricate analysis that would um that would pass muster in the journals uh in the academic journals so you know not knowing how hard it would be to build software for a really close content analysis so you know for this

like very rigorous conciliency um I started doing it and uh you know a few years later we actually had it working so um but that project we're able to look at what the the police and the cities and the protesters are doing at the Strategic level uh and the operational level and the Tactical level so like what is the strategy of their campaign what is the operation that they're leading this weekend and what's the goal of it and what what's involved how many people are showing up is it a rally is it a demonstration is it a March what is it and then what's actually happening within that event and looking at um everything happening at those three

levels moving through time across all of these actors um it's like they're very intricate kind of event history models but we're also using uh hidden Markov models to identify sequences of interaction and like particular decision points in those sequences that tend to lead to violence um or addiction or anything in between so we can with this project like with the public editor project like like with the detection um we end up involving the public and we ended up using these very transparent methods where lots of people can get involved and they can start to understand like how do we do social science wait I can actually participate in this

um wait I can actually look into every aspect of this project and understand how the data get processed and analyzed and how it produces conclusions it does so um this is like you know beyond beyond just creating the tools uh for scientists to understand we're trying to we we are succeeding I think in creating tools that give a member of the public just a curious member of the public a way to be involved that's actually helpful to the project and actually enrolls them in just the notion that we as humans can understand ourselves and once we understand ourselves even as we

interact with these complex and fraught situations um we can actually find potential Solutions and then we can take the next steps of moving forward on the solutions so um but that's probably enough on that project because I want to make sure I talk about same page which is uh kind of the way the way forward yeah uh well let's get into it yeah so um I'm I'm kind of gonna gloss over there's a there's a project that's ready to go where we uh do um we take government documents and we parse them and database them and make them easily queryable um and that that one's pretty fun and we

kind of launched that one with the social science research Council a few years back um but it was kind of a soft launch we need to get more marketing behind it um but it's that was kind of our first foray into governance you know like how can we make sure that we can watch our Governors watch our elected officials but the thing that I'm most excited about right now is just moving to direct democracy tools um we can all you know right now if we want to if we want to have a say in policy making what do we wait every four years and then these like these two people who whose job it is to fight each other uh verbally you know they get to

fight about the policy once every four years and then they're actually fighting about 50 policies at a time but we only need to vote on one person so all those with the policy dimensions are boiled down to a single personal human Dimension it's it's not I you know I don't want to bag on representative democracy or the founding fathers or too much because it was way better than what than than kind of the the monarchies that were prevailing at the time but we can do so much better we can do so much create systems where the people who are interested and willing to spend some time being deliberative can actually liberate directly about the policy choices that are gonna going to affect many of us so

um we at goodly Labs back in you know I first started uh I first started drawing stuff on napkins in 2007 for online governance um and I looked around and realized like oh I'm I'm not the first person to have an idea online governance um and so found some some great projects like um uh David Price's debate graph out of the UK and you know it was kind of poking around with that uh Mark Klein has the deliberatorium at MIT um and I think you know after there were several that we that I was looking at at the time

um there's a little group of us at Berkeley Institute for data science looking into this and came away with the idea that Mark Klein's deliberatorium probably Nails the most important aspect of this um which is what I call deliberative grammar like what are the what are the kind of units of human statement or communication that need to exist um in order for a community at various scales to talk about what are what are their issues what are their problems to solve what are potential answers or for those issues or solutions to those problems

um what what's uh what are Arguments for against those different solutions what's the evidence for against those arguments no there's like with the right kind of deliberative grammar you have something that is computationally traversable you have something that can be visualized as a space of deliberation you have something that you can easily build um input output interfaces into so that people can contribute and this as well you know like with public editor it'll be a system where um the more people contribute the more they learn how to be an effective contributor and kind of the more weight they have into the system but a fundamentally open system you know like

anyone can come into it and we want to we're building these um and I say these because we actually want to we want to build it in stages uh where the first the first product is kind of like allows you to split a bill with your friends at after dinner like let's just tackle something easy small group like small delivery um but start to provide these tools for people to work in even larger groups uh on more consequential decisions and kind of train up the world into doing deliberation um and of course a very obvious a customer Community for this sort of

Step um our doubts you know there's like how many Dows now and how many of them want to be um why Democratic but you know the forms of governance that we know from our a monthly hierarchical you know often there's like yeah I don't I don't see this too um to judge anybody um yeah goodly Labs right now we're in the process of dawification like we we want to turn in that is non-hierarchical um but for the beginning you know this is like a lot of a lot of my vision and

the ability to you know Inspire volunteers and get people uh professors and grad students and Industry folks working on these different projects um but yeah with with fan page we get to the place where we remember the tooling to bring things into alignment and even the tooling to discuss what constitutes human alignment no I'm so really excited about that we're also doing it in a pretty interesting way in that we've come across other teams who are building similar or have similar Ambitions and every last one of these teams that we've come across has really smart people who

are really good-hearted you know if you're spending your time trying to build tools for human governance and Alignment you're probably pro-social right so um we've all kind of talked to each other and just realized like you know you're trying to do it slightly different than we are but that's not a bad idea I want to know works or not give a song I agree with you I think that plurality of like multiple people approaching the problem in different ways is like a real feature uh and not a bug of the movement it creates anti-fragility in in how the ecosystem approaches things in the thing I love about web3 is composability where you can work side by side and and even like hand off uh scope to each other in

between the different modules so a big yes and to multiple people working and uh in in solving the same problem but in different ways complementary ways yeah so we're we're you know we're a couple of meetings into like pulling together as this incubation cohort we're calling calling it The governance Garden but it's um yeah the idea is that we'll just we'll do rounds of r d and then kind of share our our findings with each other um and it you know I have to admit from the beginning it may be that after a couple of rounds of r d I realized like Oh My My hypotheses weren't as good as this other team at metagov or um you know this this new team that's

that's coming out of like the AI objectives Institute or something like that um and that's okay that's okay because the the Omni win is still totally on the table there's just like if we do this the old way where each team kind of does its work in the corner in the dark and we each kind of like compete for for funding there's a high chance I think that we as a as a as a community we end up on some local maximum it's like yeah we pulled together some tools and they're kind of okay um and then that's we were kind of stuck with that for 20 or 30 years when we could have had something way better so um I think that this project the same

page which is just one team within the governance Garden I think this is some of the most important work that we're doing that we will ever do um and I'm really eager to get moving on it and then but but I also gonna get the public outer project scaled and out the door uh so yeah there's a there's a lot going on there's way more work to do then goodly Labs could ever do so we are big proponents of having a large ecosystem of orgs doing this sort of stuff like it's someone could look at it and say that we're trying to help competitors and it's like okay fine but I mean competing for what to like to do good

stuff in the world there's so much to be done we're also we've been watching this big uh Innovation engine this big pro-social Technology Innovation engine in Atlanta um we've got like 30 different organizations in and around Atlanta uh to build this government-funded uh Innovation engine that kind of starts in the historically black colleges and universities and pulls in um Engineers mid-career folks from around the Atlanta region uh and we've got the Martin Luther King Center as like a user testing community and the

rainbow push Coalition the Atlanta mayor's office and their Workforce Development all of these hooks have kind of signed letters of um of commitment to this to this big proposal that will bring in hundreds of millions of government dollars to create more and more pro-social Tech uh and this is like AI Tech it's decentralized uh it's distributed Ledger stuff and it's internet communication stuff so I'm really I'm really excited there's so much work to be done I'm extremely excited to be here talking with you and with your audience because I know that they're the people who can actually do the work and and have some some inkling to do it um but I'm feeling a little

scatterbrained talking to you frankly because I'm just like you know but there's a million things to do and I have to I have to get everyone doing all of it somehow uh because yeah although I I feel it I I mean I place multiple bets at a time also and it's about going broad but also going deep enough and um yeah so I don't I don't think you're coming off as scattered I think that you're playing placing multiple bets and um and and I think that that's it's just uh what I respect is people who have a strong theoretical understanding of what they're doing movement that they're building in their time their their place in space-time um in you know like human social progress space time but also like are placing deep bets in in multiple places so I'm wondering if we could shift back

into the theoretical about how this all aligns for you I mean we only have 10 minutes left that can go a few minutes over I don't want to do that to you but um I'm clearly yeah what is bull well you know I like for me the the sort of like nutcrap notice um Building pro-social Technology and this idea that AI alignment which everyone is buzzing about requires human alignment um and so you know how do we build pro-social technology that helps us evolve with these with AI movement happening but also you know the trends of social media and web 2 continuing to play out as well you know how does what is your vision for how this all comes together and I guess another way of asking this would be if you're maximally successful in five or ten years what

does the world look like I think that there's an important piece of the human alignment challenge that comes to us from Human Experience certainly in the 20th century but going back many centuries and that threat is around legitimacy um what makes the what makes any Authority legitimate and I don't mean like a single dude I don't know like some single old white dude or something but I mean um even something like the public editor system the collective intelligence what gives it Authority what what makes it legitimate um and if we're going to solve human alignment if we're ever going to get to

a place where we're like yes we more or less at least partially solve human alignment that solution is going to be something that um a large majority of people look at it and say that is legitimate and for it to be legitimate it's going to have to be transparent it's going to have to include a lot of people from a lot of perspectives um you know we don't say something is legitimate because uh some AI Lord um created you know a thousand AI Bots and got together 10 000 of his human friends and then you know took over California and we're like okay that's legitimate just because of power like we

don't even as as horrible as Western Civilization is as you know as much as we've killed each other legitimacy is not won by Brute Force at some point you have to have some some philosophy that aligns with what everyday humans imagine to be like kindness or caring uh you know it comes back to the family like our political economies have to somehow be at least in to some alignment with the notion that we care for each other and we take care of each other so these human alignment systems the way that I see them working

uh most likely is through big online Network communities um I think it I think it could be Dows uh I I think that's probably the vehicle for this um because representative democracies are not going to change I don't I don't think they're going to change I don't think the current system [Music] um is going to just wake up one day and say we should give away our power uh to a bunch of Internet citizens who have been talking online over the last couple years about how we should spend

the national budget like it's not it's not going to happen that way I think what's going to happen instead is that we we build the tools for human alignment and the Dow communities become richer denser more populous um and more consequential where more and more Dows are performing for people the way that the Social Democratic state was you know supposed to perform during the 20th century so our public goods are going to be managed increasingly by Dows at the same time that we have like A continuing

kind of gridlock of those 20th century representative democracies and I think more and more of the responsibility uh for human alignment and for human well-being will shift over to those dows the question is in my mind is like can we get that going fast enough that we can prove to ourselves that we can actually kind of govern and parent and care for ourselves better than the AI will be able to to me that's kind of like the the historical race that we're in against the AI is it's not about

um like will the AI wipe us out I think that's that's possible I don't wanna I don't want to like discard all of the doomshares um in a single statement but I think the real issue is like will we maintain responsibility for ourselves will we prove to ourselves that we can be responsible for ourselves and govern ourselves to take care of ourselves at scale [Music] um because once we lose that I don't I don't know how we I don't know maybe maybe the AI takes over for 50 years and then slowly teaches us how to be responsible and gives us responsibility gives it back yeah seems super plausible um yeah yeah

especially if there can be some and who's in control of the AI it just it we don't see many instances of some entity or person controlling a lot of people and then just giving them power well it just doesn't happen yeah yeah giving up power is um I mean I guess you know in a way that's a feature of democracies as term limits and everything like that but um would be unprecedented in a in a sort of AI world but um let me try to say back to you what I just heard and and you know I can compare and contrast how we how we talk about these things so um there's already an alignment problem with the seven billion humans on the

planet that can't figure out how to align how they spend resources and how they interact with each other and because of the Rival risk Dynamics at that scale um we have misalignment and then AI just raises the stakes by a thousand X intelligent agents are now out there and so the question is what is the way in which we can solve for alignment first with the humans but also the AIS and by the way there's exponential Tech that and AI is just one of them so the it's getting harder and higher Stakes how do we create legitimacy there's multiple ways of creating legitimacy the best ones are legitimacyback process by participation by awareness by performance and so um building tools that are pro-social

technology technologically that the humans can use eventually helps the AI adopt tail more pro-social path from there and um although we haven't even solved alignment for the humans at least we can start building the muscle and the coordination mechanisms to do that and you know maybe there's a path of of meat space governance to decentralize governance via dowels and that evolves and then eventually it's something that the AI participates in that's what I heard but maybe I'm projecting a little bit of a green pill back on you I think we have a pretty good mind meld um yeah I I think we're we're seeing this possibility and you know frankly I we need a lot more people to see that

possibility because it's yeah that's why we podcast or hear the possibility yeah yeah well um I think that we're just coming up on time so I'm glad that we feel like we've gotten to the Mind meld moment is there anything I didn't ask that you want to say as we as we wrap up the show and um you know maybe we should leave a call to action for the listeners to check out goodly Labs or any of the projects that you're involved in yeah absolutely I would love for people to check out what we're up to um you know we're in a moment where like I said there are three projects ready to scale and all that we need for that is as volunteers like volunteers who are doing little micro tests out on

the edge of these collective intelligence systems but also um volunteers who you know I know a lot of people uh are unemployed at the moment in San Francisco uh and have really good engineering skills maybe do a tour with goodly labs for for six months or or so working on some of our stuff we gotta launched a very good hero we're always in in need of hopefully not always but we are currently we need to launch this stuff a lot of these tools we want to provide as public goods but then we kind of have some premium uh services that we can layer on top so

they they can be sustainable pretty easily I mean just imagine like a lot of people probably aren't an AI detector that they can run their their kids I want to view the Dow governance uh proposals and be like this person is using an unhottom and that's an appeal to Authority this fact is wrong so uh yeah I'll bootstrap your tool yeah I remember we we think that we think things are going to be like sustainable once they're once they're up and going at scale but yeah that's that's the big challenge right now is like let's just get out there in the world we also we need to figure out ways so people probably have better marketing mind than

I do like how do we make this more fun how do we make this more like a movement um one of the things I'm I'm here at my my buddy's place and we're going to work on an AI Genie that will kind of uh amplify and call out some of the that's in the news to help people understand a that these generators are kind of also convenient in every case um but B that there is a lot of manipulation and that uh I'll like this this that we can all kind of confront together and and we're into a shared reality uh so we're we're trying to imagine like fun ways of uh

getting more people involved but this does need to be you know I think I think if we're gonna write the ship of the human species and and take responsibility for ourselves we need a lot of people kind of getting involved in little ways from time to time on this sort of stuff like and that requires a little bit of a shift in Consciousness people realizing not just a consumer I'm not just someone who who punches a ballot every four years like I am actually someone uh concept of consequence who can shape my own society and I get involved with some of this Civic engagement Tech like I can actually be Shifting the way that

this understands itself and and and changes itself and improves itself as we go so um yeah more more power to that uh from anyone who for has you know some energy to put into it yeah there's a word I recently learned called agentic which is basically someone who has uh the power to control their own goals actions and Destiny and I thought that was just such a profound word to concisely wrap up um the people that are building the movement with uh uh that I love working with most are uh agentic and um you know it takes a certain amount of deprogramming yourself after being a kid and being at school and you know sometimes there's like circumstances that prevent you from being agentic and

it's certainly a position or privilege to be agentic in the world but uh you know having the agency to uh to help move the world towards more more pro-social thing is I think one of the highest things we can spend our our time on so gives me a lot of Hope but yeah I I you're giving me chills because this is like this is the the term of the moment for me as well is like how can we how can we increase human agency this is what's really valuable and special about humans you know there are people in San Francisco now who are saying like you know if if we just have an AI world and then at some point the humans don't exist anymore that's okay and I'm like what what about human agents what about the fact that we can feel what we feel

and experience uh you know a sun set and we some delicious meal and create music and the AI is not going to have that experience and it's not going to have that agency it's going to have this deterministic agency that is um you know what what a cold world without humans so yeah we need to we need to generous human agency and we need to strengthen it we need to make it deeper and more powerful and Link individual agencies together through some of these social tools so that we can have a super intelligence about ourselves that that makes us both as individuals and as

social beings just more and more agent more and more powerful more and more capable caring for ourselves well I think that's a great place to end it and um you know I appreciate what your the work that you're doing in the theory space and also in the practice space Brigham so thanks so much for coming on the green pill pod and um can't wait to follow your progress of these products as they launch and and your career Arc as a growth social technology Creator so thanks again thank you thanks to the audience and Kevin for all that you're doing because this is a is a big movement and um you're kind of a Shining Light in it I'm really honored to be here with you today thank you yeah thanks so much likewise to you you smoke

[Music] [Music]