Summoning Ether's Phoenix with Karl Floersch

Transcript

foreign [Music] ation on the podcast today I have Carl flourish of optimism and we are talking about retroactive public goods funding which is basically the idea that public goods are important to reward and it's easier to reward them by looking retroactively at what was valuable as opposed to speculatively looking forward at proactively what could be valuable and so we talk about retractive public goods funding we talk about the myths of ethers Phoenix which is a Institute that could be summoned in the future and could reward people who contribute to public goods so basically it's this reflexive cycle that if we if we summon ethers Phoenix that means rewarding

people who contribute to public goods and creating a future that's more likely to be supporting public goods Carl is really good at articulating this Mythos of ethers Phoenix and how ether's ultimate impact on the world could be Distributing public goods and coordination amongst the world so on the podcast we talk about retroactive public goods funding and how in a way that is summoned Phoenix and we talk about the op stock how that supports that we talk about practical pluralism how things like optimism and dial drops and protocol Guild and git coin give it CLR fund and all these other public goods funding entities relate to each other and then we talk a little bit towards the end of the episode about how AI alignment could be front run by creating more human alignment so basically by

doing public goods funding experiments and aligning the humans that makes it easier for the AIS to be aligned but they enter the chat so a really fun episode with one of my favorite and most charismatic and technically talented builders in the space Carl flourish and I think you're really going to enjoy it Athena Dao is addressing the underfunded areas of women's health such as menopause and endometriosis Athena is a Community Driven doubt giving women a say in decision making and funding the scientific breakthroughs that impact their lives with cutting-edge technology like molecules IP nft framework they invest in research IP that ensures exclusive licensing rights Athena Dao is raised awareness to women's health research working alongside existing organizations so join Athena Dao today

to help change the landscape of women's health research what's up Carl how you doing what's up how's it going pretty good thanks for uh having me back yeah I'm optimistic about this episode hey hey um all right so let's go let's Dive In The Deep End what is ethers Phoenix remind us all right so ethers Phoenix ethers Phoenix is the ultimate hero I almost sometimes call it the rationalist God it is the uh future state where we live imagine a future state where cooperation is the optimal strategy okay you've got that in your head well

imagine you're in that future State and you want to pull us closer towards it as in as Humanity you want to bring us all closer to a world where we are all cooperating where it's all good and happy well what would you do well you might reward the people who started getting us to that future earlier the earlier you start helping us build towards that future of goodness public goods coordination alignment with Humanity the sooner you start working towards that the more you will be rewarded in that Utopia so that is ethics Phoenix well you might recognize a similar uh kind of reasoning

from the less wrong Community which is which came up with this concept of roku's Basilisk it is kind of that is kind of the opposite version of ethers Phoenix it's like the evil AI that will punish you if you do not help it come into being so what if we add another system another sort of system like this which instead of punishing you for get not bringing it to uh to about you have that thing reward you for bringing about a positive aligned future this the reason why this is important is because we can actually visualize that that future exists and if we work towards it

now not only will we get to live in it in the future but the earlier you start contributing the more impact you have and therefore the more you will be rewarded in that lifetime so this is really what I see as the ultimate kind of goal and thing to strive for and it's it's kind of crazy because we do exist in a time where I am as a kind of the I feel rationalist I feel materialist I feel down to earth that kind of thing and yet I genuinely believe that there exists a kind of heavenly Utopia which we can materially realize

um and so I'm uh I'm definitely a devotee of ethos Phoenix um and so we did an episode a year ago where we talked about euthris Phoenix and I they're in I remember from that and also you just saying so Rocco's basilics is an evil AI that will exist in the future will punish people who know about it and don't summon it etherus Phoenix is sort of like the opposite uh an angel or or some sort of entity in the future that will reward people who contribute to public goods and will reward you if you summon it so it's like an inverse recursive cycle yes exactly so it's fun to uh talk about this a year later when uh when you first talked about told me about Ripa's basketball so

I was like oh that's like a far-off future thing and now ai is just everywhere A year later so um curious to talk about Ai and coordination as the episode goes on and Rocco's basilis feels a little bit more visceral this year than it did when we recorded an episode last year yeah exactly and there is a secret undertone in terms of uh this whole public goods funding uh incentive alignment using blockchains and decentralized protocols to align incentives with human values there is actually this underlying theme that has kind of existed in the subtext for many years which is this concept of okay the world is really changing and we need

systems which promote human values in and defend human values in a much more uh systematic and uh intentional way and that has actually been I I at least for for us building out optimism and and various other folks in the you know the ethereum community this is definitely a thing that is bubbling under the surface which is like hey there are major changes in technology in humanity that are going to radically shift the pat balance of power and right now with decentralized protocols the end of the day decentralization is about power redistribution it's about giving having a more flat distribution of power and

making sure that the right participants are the ones that are getting rewarded and so there are these uncanny connections um that are happening and are becoming more and more prescient as the you know AI advances start to materialize even though we knew about it for quite some time well um I want to get deep and dark and weird on the philosophy here um in which I would say I know that you and I will do but um first I want to go tangible real hard and um I want to talk about retroactive public goods funding round two which is around it optimism brand that uh rewarded people in the ethereum ecosystem that supported public goods especially around the optimism ecosystem and um could you tell us how

that went and what you're excited about with retroactive public goods funding because I feel like that's the the tangible example of of ethers Phoenix being summoned in so many words yes exactly this is this is our attempt to really materialize ethos Phoenix candidly so retroactive public goods funding round two so this is the second round of retroactive public goods funding first off let me talk about what retroactive public goods funding is in the first place so it is essentially an idea which is very similar to ethics Phoenix where we are taking a step and we say okay we are going to look back at history and we will say who has provided the most impact for who has built out

the most valuable public goods this is a thing that we can do that we do multiple times at over and over again we're on round two but the idea is to go to round three round four to infinity and then we give rewards tokens some kind of you know uh reward to the people who built out these public goods so very tangibly the the idea is that by doing this repeatedly over and over again and establishing it as a norm people can start building public goods which actually further the cause as in they they are they are intentionally when they sit down and they're like okay what am I gonna do

today they decide oh I am going to build out public goods in the hopes that I will eventually be retroactively rewarded for my contributions this is very very similar in fact it is pretty much the same thing as ethers Phoenix where what ethers Phoenix is essentially the public goods funding the retroactive public goods funding mechanism way way into the future because we're saying hey if you help create a world that is funding public goods that is aligned with Humanity then you will be rewarded retroactively for your contributions we might reward you uh along the way right now with these recent rounds but eventually you the

rewards will be greatly Amplified given you know assuming the the mission is complete and also you'll live in Utopia so that's great so anyway the the more tangibly because I keep getting drawn to ether's Phoenix I guess it's the intent but um more tangibly the first round we distributed um uh uh of like a million dollars at the time of eth um this round we distributed uh 10 million op you can can look at all of the recipients they were all of these ethereum and optimism community members who contributed things to the op stack that was the stated uh stated goal so one of the things that we learned from

round one is we needed to actually scope the round we need to do something called impact scoping where you figure out okay I'm going to make sure that we were going to run this round we're going to reward people but if we start rewarding people in for just impact they've had generally that is going to be incredibly diffuse so we need to figure out what is the epicenter of impact that we are going to measure for this round we scoped it around the op stack for round two and we distributed 10 million op to various open source projects uh educational resources um tooling that all of these different things in the community there were the

people who actually decided who would get the tokens were um one around 100 badge holders I think it was actually 70 who actually voted but around 70 or 100 badge holders they decided they picked a distribution and they distributed out to those tokens and so this is uh you know we just ran the round it was incredibly fun incredibly cool and um you know we were talking about running this round I think a year ago um on this on this podcast and so I'm excited that that it went down yeah so what I love about retroactive public goods funding is that it feels like it's the actual tangible implementation of ethers

Phoenix and it's proving out that we can retroactively reward public goods and so um by building the muscle of allocating Capital to public goods we proved that it's possible the world's agents in this case open source software developers can start to work on public goods instead of you know working on an Ico or whatever their private private game would be so like the author's Phoenix philosophy is one thing but then there's like the tangible manifestation of it with retracted public goods funding and um I think that's what's neat is that this isn't just a philosophy it's something that's being put into practice exactly exactly and we do genuinely hope to get to the point where you know these retroactive public goods funding rounds

are being distributed based on a large set of human being preferences for for context uh the optimism governance uh system consists of these two houses there's you know the op the token holder house the token house and the citizens house which is based on human identity and this is incredibly intentional because if we're going to materialize ethers Phoenix we need to be giving representation to the preference set that we care about and and really realistically in the in the long long future we hear about human beings Distributing out the ethers Phoenix Rewards or at least being the ones who are uh whose

preferences are being optimized by the by the system uh I feel like a year ago that would have been like this abstract point to me but now everyone's talking about Ai and AI alignment and so I'm excited to go there on this episode yeah well the and even before we go like in a full-on crazy um mask off you know round three we're already starting to build these conceptual Frameworks so I think there's actually really interesting progression of like okay we had this crazy idea where we're testing it out and we're getting people involved and then we're getting empirical data to learn about how to make it better um so the first round to the second round one of the things that we learned

was that we're going to be Distributing out uh Rewards tokens we need to be measuring the impact that people have and we want to be giving more rewards to people who have impact and that have not received requisite profit from their impact in other words the definition of public goods on the surface is not super well defined right there's non-excludable non-rival risk Goods which you know information is like the typical example where you can copy and paste a file and it costs nothing and you can share information that's great then there's another definition which is good for the public

um which is the kind of original you know old-fashioned definition and the reality is that we kind of need to figure out a way to balance both of these uh definitions and realistically we need to be measuring what is good for the public and a measurement of what is good for the public is going to come down to a tangible number and so we realized that we needed to actually uh think about not only you know how much impact these these projects are having um but what is the current how much have they been rewarded already for their impact because it's not this is not to say that the people who work on you know highly profitable businesses are not having

incredibly high impact that's in fact that's kind of what the whole point of capitalism is that a lot of the times when you're maximizing profit you're also maximizing impact but we're just seeing that there's a there's a kind of a break of some things that have really high impact are not getting maximized or not getting profit and that's and that's the the classic you know tragedy of the commons and public goods funding and so this is this is the the learning that we had from round one to round two it's like okay we need to be really taking a system a systematic approach to measuring quantifying impact and quantifying you know some uh form like light uh uh High grained um form of profit and then Distributing

rewards accordingly and additionally we realized that we needed to be a little bit clear about what the Impact scope is and that's why we went from a kind of just broad oh optimism and ethereum Community to things that impact the op stack but something that we have realized from round two to round three is that we actually want to not only be maximize uh rewarding think you know contributions to the ob stack I think we're going to talk about this soon but at least these and these plans are probably not fully finalized but the current thinking is okay we need to come up with an abstract concept of what is good for the collective and then we can

figure out what the kind of valuable things are to the collective and then reward those things because the op stack is not the only thing that is good for the collective there's a lot of things that are good for the collective a lot of public goods that benefit the group and we need to be measuring and uh uh quantifying that so we're basically getting getting tangible uh metrics and getting tangible results in terms of what we are trying to maximize and reward um as we progress and as we do more more rounds and and we're getting you know less wrong um with our with with each round right yeah well what I heard there that I think is incredible is that there was

um I think it was 20 million op in retroactive uh 10 million oh yeah for this round yeah yeah these are sizable amounts and then you all are being very sort of conscious about how you sculpt the design of the experiments which I think is is cool and the net result will hopefully be that if you're someone in the ecosystem you have a rational incentive to contribute to public goods that is higher than than what you had before so it takes us away from our own private incentives into doing what's good for the collective and um there's an overlap between the optimism public goods and what's good for optimism what I'm excited about is how this could scale out to a whole world of public goods go from optimism to

ethereum to open source software to there's a whole plurality of public goods that I could just rattle off in the world like uh things like roads and bridges in the real world and um uh like economic infrastructure client a livable climate pandemic like pandemic response herd immunity and um you know AI alignment is arguably a public good and so there's like this whole plurality of public goods and I feel like you're just starting with a very small sliver of of creating the leverage to get to to to get out to all the world's public goods Yeah a hundred percent and what you're kind of getting at is the secret sauce of crypto in general and that is the social

scalability of decentralized protocols and the fact is that if we build out the right incentive structures and the right protocols then we can actually no matter what background you're from no matter where you know where you live you can look at the actual functioning the code and the rewards and and how the thing fits together the the decentral the protocol how it fits together you can look at that you can analyze it and you can say hey this is actually something that I would benefit from and therefore I would like to participate in and the the magic of CR of crypto which I think in general like maybe is is often for you know not not appreciated is that

we have the ability to create these decentralized institutions that can really help Elevate humanity and Elevate individual people and that we can that we can actually crypto while it's seen as this like just the epicenter of capital interests and you know money and all you know all of the that that kind of thing actually I am by far most excited about the opportunity to make sure that the protocols that we live in the society that we live in is governed by protocols that

reflect my values as a human being and also reflect all the other human beings values and that we're not like this whole web 3 thing the whole point of it should really be elevating and giving a platform to humanity and I think that the the social scalability of of these decentralized protocols being able to analyze them independently and come to your own conclusion that this is actually going to be in my interest that property is so incredibly special and it really it that is how we can go and grow the systems Beyond you know what any uh Corporation or even government could could provide so I I'm I'm really

excited for you know decentralized protocols for Humanity yeah yeah same um if anyone's watching on video you can look over my shoulder right here and you can see I've got a knee through Phoenix up on the wall behind me and I recorded every podcast episode in front of it so um I'm drinking the Kool-Aid for sure oh it's so nice well uh one thing I want to ask you uh that has changed since last year is that there's now more than one retract of public goods funding program out there so Dow drops uh recently did a red track to public goods funding round and I don't want to put you on the spot but what do you think of other people who are doing rpgf type experiments I am super excited about all of the different

retroactive public goods funding experiments okay I will confess that um my background is I grew up my mom was very uh Hindu and my dad is this uh kind of skeptic uh you know intellectual type um and so I had this weird upbringing where I kind of got a very deep like science first approach from my dad and then I got this Universal Consciousness approach from my mom and so when I see people doing retract to public goods funding uh experiments in in isolation I actually kind of see those as just pieces of the broader you know the

broader puzzle or the broader picture towards ethos Phoenix in that these aren't necessarily if we are all optimizing if we all have the same vision for what the few future should look like roughly and if we believe that that vision is is one where human beings have the ability to you know are are elevated and and and our our economic systems are aligned with our interests and that goes for you know they'll hold all facets and I keep we keep touching on this AI thing but um then if we're all aligned on that being the long-term Vision then each one of these experiments are kind of bringing us

closer and I do actually believe that there is a convergence and I of course it's not going to exist today absolutely not um and I'm a big fan of of of plurality in terms of getting us getting us there but then there's always in in the context of a like a plural economy if you take out if you if you go one step elevated you end up with a kind of single ecosystem or system and so I do really think that like there's in some sense we're all distinct in terms of our retroactive public goods funding rounds but then in some sense we're all as long as they're we're aligned with the vision moving towards the same the same uh uh end State uh yeah I mean I I think that

um plurality creates anti-fragility right because if you've got multiple clients that are implementing similar protocols then you don't have to rely on one company or um one implementation and this is one of the things I think the ethereum network has done really well is that there's now half a dozen consensus clients that you can run and we're getting there with the execution clients as well and um you know this idea that because no client has a super majority of the network you don't you're not captured by that team and also there's a bug in that client there's no fault in the ethereum network there and I almost think taking that philosophy of of pluralism client diversity and bringing it up the stack to other things that are existentially

important for Humanity like public goods funding is a good thing and so it takes entrepreneurs like me with what my work I've done at Bitcoin and you would optimism the Dow drops team and Griff from gibbith and Oren from CLR fund to kind of represent that it's it's about public goods in the mission and we're all just creating vessels to get there and not pluralism I think is important and it's something that's different than uh then then we had in web 2 where we would all elbow each other out for market share so I'm really all about practical pluralism for for many many reasons yeah oh so many things to respond to there so first off fully agree we are we I mean even even just in optimism in the L2 space we are already we have actually

a um alternative client uh um consensus client implementation for for the you know optimism protocol and also for the op stack and also a alternative execution client um for the op stack and these are these are clients that teams that are not the original you know optimism crew uh actually created so it's really it's really it's absolutely critical that we are I think that in general what you're talking about with the ability for distinct groups of people to collaborate in a genuine way that is the superpower of web3 that we are able to have this

concept that like okay if you succeed in your in your you know in your area that and I succeed in my area we will actually benefit each benefit from this from that each other's success and this is this is in contrast to a lot of the as you say a lot of the web 2 space where it you know if one Web Two company succeeds in one vertical or one one you know one part of the value chain and this is why we actually talk about value chains then another company in the value chain might actually have to start paying fees to that company and you know eventually uh succeed and maybe get like merged into that company right this is this is a constant struggle and the web

2 space and I think that there's it's really interesting because it kind of Blends the two uh sides of the two like left and right sides where you have this ability to have a free market with uh collectivist um you know principles and and that really is a superpower and I I appreciate all of the the like public goods folks in the web 3 space who like carry on that torch um as we move forward and as as uh as as web3 evolves um because we are all building towards that long-term that long-term goal and along the way it's it really is I think it is actually a competitive advantage

that we are able to coordinate and work together to build that long-term goal yeah totally that modularity where we can each do one thing and do it well but um it's kind of like super modularity because they all of the different things complement to each other and and so I Envision a future in which let's say in five or ten years that uh if you're working on open source software or any other type of public good then um you'll be getting rewards from hopefully get coin hopefully optimism hopefully dial drops hopefully giveth and hopefully see our CLR fund maybe protocol Guild if you're working on that specific public good and so I I think that that plurality and the multiple income stream

is is going to be really incentive shifting for the world's agents to have them work on public goods but I also just like the idea that there's gonna be multiple mechanisms and it's a relief off my shoulders as the quadratic funding guy that quadratic funding doesn't have to be everything to everyone it just needs to do one thing and do it well which is measure do your peers respect your work whereas retroactive public goods funding can be like a different mechanism it's the sum of all those things together that is is gonna hopefully mean that people can shift their incentives towards working on public goods in the future and that's how we create human alignment uh which could be a foundation for alignment of say smarter agents down the line Maybe so any response to that I mean I am fully in agreement and I

guess okay I'll start bridging a little bit I think that there is a really interesting both there is plurality and there is unification at the same time in these systems in these complex systems so so here's a here's an example of two two different problems that I think um are are really interesting to separate as opposed to tightly coupled them there's one thing which is uh and there's one thing which is rewarding public goods creators that is one that is and and that will come from various mechanisms across across the ecosystem there's another thing which is measuring utility of the public goods and that is

in some in a lot of ways really separate from the reward mechanism because what you want to really do is you want to First measure utility and measure the the the the impact that these that these projects have to the group you know some group of people and then your various reward mechanisms use that information to incentivize the high impact projects right the normal the normal way that this happens in the you know in uh private for private Goods is impact is measured as how many people are lining up to buy your coffees at the coffee shop that is a very straightforward way to measure impact but the classic

challenge with all of governance is pretty much how do we fund public goods when we don't have that obvious price signal that private Goods provide us and so the way that we need to the way that we can do that is we can actually create an ecosystem of people who are trying to measure the impact of various public goods and you can create that entirely independently right again the modularity piece you can create that entirely independently of the various mechanisms for rewarding the impact and me and and consuming that data so we actually don't even have it's not even just one monolithic retroactive public goods funding system that has different you

know segments in it no this is actually the the modularity extends even within the retroactive public goods funding realm and then and then there's even more with Hyper certs and incentivizing proactive you know work to be done on that this also so that that's its own its own thing but no I think that as we talk about it you start to see the alignment even just kind of naturally emerge between the problems that we are having with aligning the broader economy with you know things like AI safety or even in the future having these you know AI agents that are that are acting in the economy directly with this Pro and and trying to make sure that that's aligned with Humanity with the problem

of all of these different Dows that are trying to optimize for you know who's having the highest impact on on on the the users and really it's on on humans um so this is yeah well I've done a couple episodes with the hypersearch folks and I think that they're going to be tickled pink that you name dropped them um curious if there's any other tools in the budding public goods ecosystem of which optimism is kind of becoming a pillar that you're really excited about I mean of course get coin right like we the the whole projects registry and all that kind of stuff I think it's it really is uh I I think that right now is actually an interesting opportunity

where we are beginning to see the ecosystem emerge but candidly the tools are unbelievably rudimentary and the way that these tools interact with each other it's right now it's not super well specified and so one of the things that we're actually trying to do you know not to go back to to some optimism stuff but gotta talk about it we're we're trying to get to a point where there is a reasonably clear definition of what this pipeline looks like in the abstract or at a high level and then communicate that well so that everyone can actually slot in and this is this is like part of the the

mentality of like building out web 3 projects it's different from building out a traditional you know system is that being able to communicate a framework but then enabling people to slot into that framework that is actually how you Empower an open source Community to really do something that is far beyond the capabilities of any traditional you know top-down orc but you need to you need to supply the framework you need to supply that Clarity um we've like talked about it like the Retro pgf protocol or you know various other things but this is like we need some framework for all of these people to to slot into so anyway I'm excited

for the tools that exist today but I'm also excited for the tools that are absolutely coming well that sort of brings me to in my mind is I was actually just thinking about Oren McMillan and sealer fund and uh and I always talk about practical pluralism and this idea that hey Bitcoin and CLR fund are actually working together on creating more public goods and just you know we're very inspired by the client diversity of the ethereum network but I was sort of thinking that hey the the ethereum network we have a specification for how the evm works and how consensus protocols and everything work and everyone works off of that same specification and we don't really have that same thing for public goods and um until you said what you said about you know setting that framework I was

thinking that it was going to be a kind of units philosophy type thing where you know unit the Unix philosophy is that you write your modules to do one thing and do them really well but you write them so that they're ready to interoperate with any yet unforeseen tool not yet written tool and they can interface with your API and that's what where you get piping when you're on the command line in Linux and so um the public goods ecosystem it feels like we don't have the main ethereum spec that's being delivered to us but we do have this sort of like emergent Unix style thing where things can interoperate with each other by planning for unforeseen inputs being built on top of that and um you know the last thing they'll say is that when we embarked on

the 24 month journey to turn Bitcoin from a company into a dow a uh uh platform centralized platform into a modular protocol it was because of that modularity we wanted things to be able to interoperate with each other that we were really after um also to get rid of all my spaghetti code that I wrote in the 1.0 days but uh so yeah do you think we're going Unix philosophy Style with public goods or you think it'll be more like they'll be the op standard for retroactive public goods funding which which one do you think will emerge oh I mean I think that really the the Unix philosophy and I okay the Unix philosophy requires a relationship between the processes and piping it needs a kind of definition of Unix in order to actually achieve the

modularity and and slice out different uh you know pieces of functionality and so actually what I think is the most likely is that we can't have a complete specification for the entire thing but we can start to agree on or form consensus on the way these what the different components are and how they can interact with each other for instance piping and that with that we can then really supercharge the you know the ecosystem and I think that this is not just applied to you know retroactive public goods funding or various other uh uh things that we're talking about here but this is just in general if you're

trying to build open source software this is the kind of uh this is the secret sauce the magic sauce of uh building out a really thriving ecosystem it's about it's this interesting balance of framework and organic growth and I think that candidly I mean the ethereum protocol is definitely more framework than organic growth itself but but it's also a good example of organic growth because the evm specification is a hugely fertile ground for organic growth and various you know smart contracts yeah I mean this is another OG Linux sort of era thing but the cathedral and the bizarre the idea that you have this

top-down Cathedrals that are intricately planned by and and then you have the bottoms up organic growth and it's uh I think Linux and open source software definitely resemble a little bit more of the bizarre and so um I'm excited to invite you back in a couple years and be like hey we called it it's actually happening look at this these tools that have developed ah yes exactly me too yeah well um I would love to to start uh talking about um I think AI alignment has has become a big topic over the last several years and um you and I were talking uh zuzolu a little bit about that and and I'm wondering if you have any takes there I mean given that you're reverse

engineered Rocco's baskets which is an evil AI that turns Humanity into its slaves uh do any takes on AI alignment and public goods and and uh things about nature yes so this is the fun thing so anyway we have been talking about there okay they're they're these little hints which are oh we need to align the global economy with human interests and oh this is what crypto is for and oh programming incentives you need to make sure that the incentives are actually good and then that begs the question what's good if we're programming incentives so this is the subtext as I said before this is the subtext for a lot of crypto candidly um and I think the ultimate purpose of

these systems is to achieve that alignment alignment not of necessarily AIS only but alignment of agents in the broadest sense whether that be corporations whether that be uh individuals whether that be Ai and the way that we kind of the way that the AI alignment space has kind of emerged at least as uh from from where I can see it right you have these two types of AI alignment which is like internal alignment and external alignment and internal alignment you're trying to make sure that the the goal that you programmed into the model is actually being optimized the one that you

intended so if they're playing a video game they're not just hacking out of the The Matrix or then there's external alignment which is making sure that the goal is aligned with Humanity so here are my hot takes you want to hear my hot takes hear my hot takes and we can go back in a couple years and see see what it is all right I think that the that external alignment is probably best defined as maximizing human utility over an infinite time frame so some notes there it does not mean that the AI must know what is good for humans over an infinite time frame that's not what I'm saying what I'm saying is that the AI must at its root it's intrinsic it's

terminal motivation it must always fall back to is this going to contribute positively or negatively to my predicted value of human utility over an infinite time frame here's an example the classic paper clip example we can go through that where we say okay we give the AI this objective to maximize paper clips so without the AI align this this terminal goal this terminal uh intrinsic goal we would see the AI go in and they start maximizing paperclip someone tries to shut it down they're like hey you're going to make me stop maximizing paper clips the whole world becomes a paper clip whatever fine but if we add in add in that layer that that kind of like similar to like

the Asimov rules we have that terminal goal which is okay in case your stated goal if it disagrees with this final thing which is hey maximize human utility over an infinite time frame it would be very simple to shut down this paperclip maximizer because the paperclip maximizer yeah it's trying to maximize paper clips but it's only maximizing paper clips as long as that is not putting in Jeopardy the ultimate utility function which is maximizing human utility over an infinite time frame so we need to encode that into all of these AI systems that should be the implicit kind of thing that is at the root they always will make sure they do that and that that way right whatever the ai's motive is it is going to be in

line with humans you know human subjective functions okay now that's one thing but that pegs another question what is good what is how would you even predict how do you know what is human utility over an infinite time frame what that doesn't that that's crazy Well turns out that we are attempting to measure it we do not want that measurement to be a subjective Vibes based thing that AIS are coming up with for themselves no we need governance protocols to measure that human utility measure it not only uh not not not into the future but we need to measure today

what do we know about the utility of human beings okay that begs the question of like what are the units how do we measure this like what is that actual what is that system well that system I think is pretty much the same thing that we need to serve as the basis for retroactive public goods funding not necessarily for the collective or some small group but the ultimate measurement of what is impactful and what is net negative we need to create that value graph for Humanity for all these people so and then we can have an actual tangible metric which is this is our

best guess of what is good for human beings then we can actually say we have a way to encode in the AI they can check they can check their work they can do something and they can see what human beings actually if they benefited from it and they can predict they can say hey this given this Oracle for human human preferences what is my action going to do to this Oracle in the future and this is exactly the same prediction as the prediction that we're talking about making where Builders are saying hey how can I maximize my impact if I want impact to equal profit I want to be profitable I want influence I want money well I need to build what human beings

want that means we measure what human beings want and we reward the things that benefit us so this is where the I think that the the worlds really do Collide in a very concrete way um where we're actually going to be building out the same infrastructure in both the public goods funding space and the AI alignment space I'm off my soapbox love it um I'm gonna let me know if this is true reductionist uh AI alignment requires human alignment the way we get human alignment is with the retroactive public goods funding public goods funding summoning you through Phoenix is that true reductionist

that I think that's legit that's it if we're going to if if we want to stand a chance to making sure to to making sure that when AI come into the scene that they have the ability that they are maximizing human utility we need systems which incentivize people to maximize human utility and maximizing human utility is public goods funding because what is a public good it is a thing that is good for the public and so these concepts are just literally the same words in different in different uh uh kind of ways utility of course has this connotation of we're

going to measure it right and some people are like you can't measure that but I would say to that just as an aside that if we don't measure something else is going to get optimized and so I want to at least do better than no measurement at all and we don't have to be perfect I don't believe that there exists a true utility or a true utility function for me or for really or for Humanity but as long as we get close then I think we'll stand a much better chance of Living Infinite lives and uh over an infinite time frame and then go into a black hole once there's a the universe heat death occurs

yeah what I um I mean I do think that like there's not one utility function but you know again this takes me back to pluralism and you know one of the things and like by the way respect to retroactive public goods funding I think this point about um like we retracted public goods funding the whole idea that it's easier to see whether or not something was good as opposed to whether or not will be good in the future that is a great unlock in measuring the utility function of public goods funding but I really want to live in a world where there's retracted public goods funding and there's quadratic funding there's quadratic voting and there's conviction voting and there's Assurance contracts and like let's do all the mechanisms will be a plurality of utility functions and let's build them for ourselves now

but hopefully when the AIS or AGI comes out there we can just say hey we this is your utility function just like take over what all the web three people were were doing and I think that that's that's kind of the goal for me I think I have a similar goal but I think I I mean you know for whatever for whatever reason I'm kind of like a a spatial I want to come up with a like a spatial diagram of where all the tools fit because I strongly agree that all the tools will exist but my intuition is that each tool is like it's like a screwdriver versus uh you know Phillips head something I don't know how to make do handy things but anyway

um the point being that they're all these slightly different things with their slightly different trade-offs and that they can actually all exist within the same ecosystem all supporting each other in various ways like to be clear retroactive public goods funding in what I was describing with AI alignment is not the same as the measurement problem or the Oracle for you know for impact that is actually different I'm kind of decoupling those two things so that they can exist and have space between them so that we can use the this Oracle as a foundation for not only the economic rewards that we're Distributing with retroactive public goods funding but even encode it into the the kind of the

the the intrinsic motivations of these of these AI systems that I think that we will need all of these tools and what I'm hoping is that we're going to be able to before we know how to build all of the tools that we will know how they relate to each other and then we can motivate just like in the Unix philosophy just in this pluralism we can motivate all these different agents to work together on their specific lanes and and then pipe all the instructions and information between them so I'm I'm that's that's my little rad set of caveats yeah that's that's really fair um I wonder if you could rattle off okay

so if this is a Unix style system of components what are all the components that that people should be building and maybe I can put it as a challenge out there for someone to create a system diagram or something like that but but off the top of your head what are the what are the what's the equivalent of cat and LS and CD and said and awk like what are the components that we need to do nice okay so there is proactive funding retroactive funding and impact measurement I would say those are the big categories of things so the proactive funding first actually utility or impact measurement we can start there so first off we need to come up with systems for measuring impact or measuring utility whatever they're

interchangeable roughly so okay that means that the participants in the system we need to figure out what is good for them what is not good for them this is kind of like defining a public good it's not a definition that is going to be like here's an English language definition no it's going to be systems which are probably realistically also AI assisted for actually measuring true impact at least to some approximation of external factors agents in the world on me whether or not that's good whether that's bad whatever that is one thing that's impact measurement and this is implied in literally all of these systems if you are going to distribute rewards retroactively you need some kind

of impact measurement system or you're implying an impact measurement system every time you start Distributing rewards because you're kind of saying you get the reward and you don't all right so that actually brings me to the retroactive side retroactive in consuming this data you're sending out rewards to things that have high impact okay but this is not cool this is clearly not enough because now you need an enormous toolkit for people to work together and actually build things that are maximizing utility and so that is where a huge number of these different systems both actually I

mean I will say the conviction voting all of the different voting schemes that can be applied to the retroactive component as well so I'm kind of abstracting away some of the Nitty Gritty details the way I connect yeah oh you're on a roll there I was gonna inject like the mechanisms that I rattled off quadratic funding uh quadratic voting direct voting um Assurance contracts like I think those are different mechanisms that in some of them are suited more to proactive funding versus retro funding that's where they slot in in my head it was going to be right yeah exactly exactly so you've got these categories of things you've got the retroactive funding you've got the proactive funding and the proactive funding is critical because if you don't have proactive

funding then you absolutely are not going to be maximizing impact and therefore there'll be nothing for these retroactive systems to fund and then you have all of these different like you know tools which will enable people to do better proactive funding better impact measurement better uh retroactive funding and this is the kind of this is I would say the the almost like the cycle of life right and and by the way okay here's I mean I'll even here's this is this is going to get a little abstract but like we can kind of to think of the agents that are maximizing human you know that are that are seeking the retroactive rewards we can kind of think of that as some sort of weird

black box or like collective intelligence and you can kind of say okay there is assist there is this system we don't know what's actually in this black box and we're defining a reward function over here that's based on this impact measurement and then you've get you get you know Rewards or penalties based on your your behavior in this in this black box and then what we're saying is okay let's create the system of rewards so that the Black Box can start more effectively building out things that maximize the uh uh the the reward systems preferences and there's actually even analogies this is this is

where I'm saying is a little even more kind of like abstract is there's analogies with like Gans where like we have these uh systems for judging the output of one model right we in AI we have these uh models which we're trying to judge the output of the AI and we're kind of tuning the AI to produce outputs that are more in line with what we want and so this is kind of like saying okay we we don't have the we don't have a model for like this collective intelligence but we still need to tune it to do what human beings want and so that's in some sense what we're doing by building out this retroactive public goods funding system by building towards ethers Phoenix we are essentially

creating the incentive structures which tune the collective intelligence of our economic agents to reward human utility um so that's like the mask off um uh description of kind of the ultimate point of what we're building with these decentralized systems and public goods funding new front-running AI alignment right now Carl I mean I believe there's Phoenix and make sure that as many people as possible recognize and by the way to be clear if you are if I'll speak to to the to the folks out there I think if you are trying to solve this problem

because you have been independently convinced that it is a ginormous problem and and that's not something I can convince you of or not convince you of but if you have been convinced by chance then I do think there are two kind of things that you can that you can do if you want to protect against the AI misalignment future that we could exist in and I think that those two things are are roughly like are you making sure are you doing internal alignment work were you doing external alignment work and if you're doing internal alignment work that is incredibly important and I do not have a whole lot of information to talk about there but if you're doing external alignment work please reconsider the uh importance of public goods funding in your mission

because in fact I do believe that these problems are one in the same yeah I'm in I think that's a beautiful place to end I think that you stuck The Landing how can people help you how can people help optimism how can people help summon ether's Phoenix I'm really excited about this future that that we're all building this practically plural public goods funding future so how can people help yes so definitely go to optimism.io um join the Discord uh you talk to us we'll help work on one of the uh at one of the companies that are in the ecosystem trying to build things build something try to get retroactive public goods funding honestly that's probably

the best way you don't even need to like ask anybody just you know just go out and do some stuff and build towards either Phoenix and the beautiful thing again is just that like if you know if we really get this thing done then it's not like we are doing different you know if you're building one part of that this project then you are building on the project of optimism and of ethers Phoenix and of all of these other you know people out in the ecosystem so I I do really think that what is critical is getting involved and I'm uh there are people out there who on optimism.io that are ready and excited to help and join and in this

journey and also of course get coin and Hyper certs and shout out to everyone else so that's it because well uh thanks so much for coming back on the podcast talk about you through Phoenix one year after the episode that we did together really excited about retroactive public goods funding and the thrill that we're creating together so hope to invite you back in 2024 and 2025 and we'll just we'll just see if there's Phoenix growing from from here so thanks for starting the meme girl hey let's go all right peace and love peace [Music]