 really excited to see all of the funds that are coming together to talk about what they're doing to talk about how to build structures for better funding mechanisms and not just funding but Optimization mechanisms. So there's the crucial difference, but I think is tremendously overlooked in our macro system worldwide nation-state based Grant funding oriented structures, which is primarily a funding system with no good feedback mechanism and no good feedback loop And therefore it cannot optimize well. So what I hope to do in this talk is to talk about Why we need optimization to be much better? What do the feedback loops represent? I'll use venture capital as an example to kind of Learn lessons from and then I'm going to talk about a bunch of the work that we're doing to build these things that we're calling network funds or blue funds the blue and green dichotomy you'll get a sense that whole like Gradient that you see at the top that I use in a bunch of talks has a meaning We're going from the science oriented part the early long-term Development of knowledge to translate all of that science into technology that we can put in products and we can use and so on So green funds work really well venture capital funds work really well We need better structures for blue funds or like these teal in between funds So the structure of the talk is this where we'll go through capital allocation as a is a distributed optimization problem and and why Capitalism works really well and why venture capital specifically works well we'll dive into the detail of What makes venture capital work? Well, and then we'll try to back into What must network capital look like how must it operate? What are the components that we need to? build into these systems to make them work so what's going on in the in the broader landscape is that there is a Infinitely dimensional, but maybe sparsely populated Space of variables to optimize at any given moment in time and maybe what he must care about is a much smaller space than that but still extremely dynamic extremely large extremely difficult to measure And extremely difficult to make predictions about and so what's going on in our broader coordination problems and broader coordination systems and this affects us every from the Local scale of like how you're operating right now as a human to how the entirety of our species is operating at a grand scale It's all about optimization systems Thank you Yes, even like the I think the new MacBooks have not yet optimized their Their new HDMI ports so the dongles probably are still good So the the optimization process that we That we have to like get to is how to allocate and organize our resources to Move through this these these surfaces that are extremely large extremely dynamic And how to do so as effectively as possible In reality like no group gets to look at the whole surface You usually only looking at a patch looking at the small area and you can maybe Reason about locally about what's going on and you can maybe summarize that information and propagate it to other people and I'm sure that all of you have had trouble convincing somebody else of something You know to be true based on all the observations you have have seen so you know how hard it is to then have chains of reasoning and chains of descriptions so imagine how hard it is to get our systems to Coordinate on some local information and aggregate that to then make decisions somewhere else about about reality This is actually not quite true in reality like you're looking at that most of the time and The reason that capitalism has worked so well over the last hundred few hundred couple hundred years Is that it is a structure that works? Well at the at various different scales in this optimization surface It's similar to to biological evolution where it enables a runtime where new organisms can emerge They need resources so they are forced to optimize if they don't optimize they will die And if they find a good sweet spot that they can grow potentially reproduce potentially continue operating and so on and if at any point they lose track of reality They'll weather and die and that's extremely crucial to this optimization process working The other thing that's crucial it is that is that it's open and permissionless that at any point a new entity can Can start anywhere in the optimization surface and try to optimize from there using local information ideally So putting this together like imagine how difficult it is to do central planning about Around a surface of this magnitude when all the parts are only looking at a local patch And then trying to convince each other to create one set of policies and apply them everywhere Even with supercomputers now even you know massive-scale AI systems will have distributed algorithms That to reason about all the complexity in a local way and apply local policies So part of what varies in the capital structure is that what looks like this to some some group might look very different in another area Of the market or another area of the world or another industry And so you have all of these different entities looking at some small patches and tons of competition in that area As the surface is changing all the time as the surface is encountering new technologies that totally change the structure Now what makes all of this work is that there is a system of message passing and value flows between all of these parties to help organize and help optimize this large large structure cool, so One of the things that I'm extremely Interested in helping optimize is this innovation chasm. It's it's about What happens between coming up with good ideas and trans translating those good ideas into technology that we can put into products and Broadly diffuse because I think that our super the superpowers that technology grants us are Or and the superpowers that science maybe illuminates this potential only get realized when you do an enormous amount of R&D Development to get it get it out there and there's a lot of like lacking optimization in that area It there's no good structure for optimization in that in that part The incentive structure on the right is kind of like the capital incentive structure And that tends to work well for whatever the optimization surface is doing but it's very Brittle when it comes to what capital values, so if the capital structures value something that Humanity doesn't really value then you get bad results And you can probably look at the state of the world today as a good example of like what happens when you have Capital systems misaligned with human values, so maybe you know rewriting the structure there and orienting capital to Maximize for what humanity does value might yield to much better outcomes on the flip side on the academic side We have all kinds of Incentive structure problems as well where it there's Academic credit and that works as a robust incentive structure, but it tends to be to create an environment that is very difficult to experiment in where you Are punished for trying things that may not work out you're punished for Publishing things that or not just publishing but but trying things that will That are potentially promising bad directions, and you're certainly not rewarded for translating any of your ideas and thoughts and and and and good Publications into technology that that creates some broad right utility So let's look at venture capital for a moment These like so-called green funds I like Right unplug and replug or is that too dangerous? Yeah, I'll try it It might be It's worth finding other times. I think it might be cable Can you live with that? Okay, right so let's talk about venture capital the Basic structure of how venture the venture capital industry industry works is this where there's you know set of there It's really a kind of service providing industry that is helping solve an optimization problem within within the broader structure there's a lot of money in the world in large in managing large Massive massive scale funds that don't have the visibility and ability to look at some local area and the optimization surface And then on the other side there's entrepreneurs and companies and builders who need capital to be able to flesh out their ideas and get things out there venture capitalists Venture capital structures are different funds that find a spot in the in the optimization surface Race capital from other investors and deploy it into startups And so they serve as like this really useful signal processing node in between to use local information to try and deploy capital And in reality, it's not just one entity there. There's a whole set of networks in there Just in terms of I won't kind of like recapitulate the success of venture capital But you know just look at some graphs of you know most of the technologies that we use today Have been you know in some way shape or form influence by venture capital and so on The the structure has this kind of what I would like to call a ladder Or staircase structure where you have companies being able to raise different amounts of funding at different stages So that you can try out things at smaller scale And if they're working if you get a good signal then you can start scaling And so this entire structure this ability of the the market to optimize for these pieces is part of what enables the venture capital Success story that you can have many different funds operating at these different stages with different perspectives and thesis about how to do it right And and in reality have thousands to tens of thousands of entities doing this to make this work really well So again, it's a signal processing problem. It's about how do you aggregate information in that massive scale optimization surface? Bring it together into a resource allocation distributed system There's a enormous amount of money flowing through it It's been very successful in producing a lot of corporations and a lot of technology And there's an enormous amount of capital in the system and there's tons of funds. There really are it's an Supervivalent ecosystem that enables new fund managers to emerge to it to build new new structures deploy capital and if they're successful And this is the key part if they're successful They tend to be able to raise larger funds and larger funds and larger funds and the moment they stop being successful they raise smaller and smaller and smaller funds or die and that's a critical part of making this whole optimization process work So I kind of like have this model for for how the picture works Which is you have a startup or a company of some sort and selling some products into a market and creating cash flow Now we could try and sell it stock in the broader public stock market But the broader public stock market doesn't have enough information gathering about that particular company to be able to make really good decisions About how to run it how to help it and so on So you end up in a structure where instead that company will Sell part of its cash flow or potential future cash flow to investors that can help it in that stage of its life And so that means think of like fractionalizing the potential future outcome. It's not the outcome today It's not the cash flow now is a potential future cash flow as sold to another market of investors the public markets and They sell that potential perspective Measure At that moment in time. And so this is really a predict what this is doing is creating a prediction market on the future cash flows of these corporations and You know, there's of course many ways to like realize that cash flow through dividends liquidations acquisitions and so on But usually it ends up being selling the stock in a public market at some point Which by the way, it's kind of a bug in for other reasons, but I won't go into that now Let's look at a VC fund itself. So VC VC fund works by funneling cash flow funneling money into companies taking some pieces from them and they themselves have a set of investors and VCs race from other investors using the same exact composable structure so they can float this trend this the transactions of a VC fund are Composable in a super neat way They you have a transaction of cash flow going out in one direction and kind of getting stock back And you have these same exact transaction on the other side to race capital the VC fund gives stock and gets cash on the other side So you have this structure where you can feed forward cash and back propagate Stock and you have this signal processing unit And the ease compose super nicely and super neatly. So you you have many fund-to-fund structures or VCs with with angel scout funds and these you know stack and stagger and so on There probably aren't that many layers But it but what I want to serve convey to you is that there's something else that we're building a lot of that looks a lot like this That I'll get to in a moment But anyway part of what's doing this is that the fund the funds upstream are much larger in size and in capital range and these these structures enable More fine-grained allocation of resources and capital across across the board And so this is what helps build this massive staircase with tons of thousands thousands of VC funds That are allocating value this way So there's a lot another thing that looks a lot like this and these are neural networks there I don't think this is an accident I think this is something fundamental about these composable units that can over time develop and help bring optimization They're not doing proper, you know feed forward and back prop in the way that neural networks do but these are kind of Bayesian reasoning type entities with an evolutionary structure encoded into it, which is precisely what you need to optimize the surface Now let's talk about network capital and running out of time So the key thing that I wanted to to kind of impart upon is that Decomposition of the venture capital system as an integrated system with that composability of the structure So we'll look at blue funds and network capital and the pieces of those and to see how how to compose something like that in To create funding structures for public goods So going back to this problem We we what we want to do is create incentive structures and funding things that can be successful in that area of the of the market Now there isn't good cash flow optimization there or this would have been a solve problem through VC So we need something else I can that we can use as a way to Derive value here The good news is that there's an enormous amount of funding that is deployed through to science and In charity and so on but the bad news is that there aren't that many funds It's it's highly centralized The further up you go the fewer entities there are so it is not a good optimization structure and Worst of all is it that it's not composable You don't have a good way to stagger funds and to propagate funding across different different To build a network of funding structures So going back to this we wanted something like this We want to have a version of this that works to do this kind of public good funding Not just funding the creation of a private cash flow as the stock market values But a funding structure that can enable the the broad production of public goods And ideally with a feedback mechanism that that enables a prediction market to form in to optimize that surface So let's look in detail at public good organizations Then they tend to kind of create a public good and they sort of give it to to the broad public and Usually what comes back in the other direction is kind of like some sense of mission accomplished some goal Has been done that is often one of the primary reasons why people do this kind of stuff. The other parts is thanks or broadly broadly recognized credit For doing this kind of good impact this some other kind of measures and so on but it's probably you know These kinds of these kinds of structures and there is some amount of fundraising that happens where those organizations can turn to a set of Donors or you know investors in the in the outcome of public goods And propagate the credit and the utility or the sense of the mission has been accomplished And so these other you know kind of things are getting propagated back But it's not very fungible and there's no good way of scaling it So what we ideally want is we want the structure kind of like this where you can we can develop a Large-scale optimization surface where anybody can go and create a new fund and you and you're enabled these signal processing units that can operate at very different scales you need fund the funds to to exist here and Crucially you need performance signals to impact the funding scales many grant programs out there including some of the largest scale grant programs in the world do not really have a a Strong feedback loop between the capital deployed in the capital invested in a process how well that turned out And feed that back into their ability to allocate more and more capital to a field You also get cases where you have very large funds trying to allocate capital in areas where in many different fields without a lot of Knowledge and a lot of signal processing and you get into like these are very really bad feedback loops We're massive amounts of money just gets deployed in poor ways So what we need is some some composable structure the ability to create a fund structure similar to VC fund That has composability so you can a fund itself can raise capital from other funds like it That it can deploy capital and can get feedback directly in that transaction We need to make that transaction itself composable, and this is what part of the problem here We don't have a good unit to sort of propagate back However, we might be able to create a unit like this So through things like you know the public funding mechanism exploration that's happening in the in the grip of space in the dial space We might be able to get to a structure here. Let me switch So so one idea and it's not clear whether this will work, but this kind of where what I think might work Is that we can what we can do is we can fractionalize the Value flow of one of these sort of a public public good funding Structures through the use of impact certificates So if we can use impact certificates to quantify the value of a particular fund as it deploys capital in an ecosystem And use that as a version of a stock certificate In a corporation, so if we can use impact certificates to create a much more Aggregatable and and evaluatable and Legible and so on structure for the and tradable structure to deal with Impact this kind of a broad-based public goods impact Then we have a component and a tool that we can then use in this transaction to trade back and forth through these Through these entities, so that's kind of what we need in these structures over here So if we can turn this So what in the VC world is a stock certificate in the public goods world? we can use an impact certificate and Part of what's required here is a broad market like the stock market that is going to value these impact certificates It's going to value the utility of those components and it's going to pay large amounts of money We're talking billions of dollars worth of capital deployed against these impact certificates If we can get that to work then we can build an optimization surface like this and an optimization system with a bunch of little Composable units that can build you know this broad larger scale network of blue funds Blue here is like that this kind of public good oriented oriented thing One of the key things here is you can do carry structures with impact certificates. You can you can have the traditional VC style Performance indicator of carry and not have an and really couple the the performance of the fund associated with the impact that you care About doing and what's really cool about impact certificates is that you can do Impact certificates associated with specific fields You can value those impact certificates with a group of people that understand that field and can know after the fact Five years later ten years later and so on what that impact was really worth So my sense is that we can do this kind of thing for open source and we can use impact certificates in these kinds of Kind of structures. There's a great talk from Pooja in the last event that goes into into a bunch of structures I think that impact certificates can be weaved through most of these We can do this in the science in R&D funding We can create as we have the traditional fund grant systems We can start flowing back impact certificates and start accumulating them We can just start doing it and start issuing them and just accumulate all these impact certificates Once we have a bunch of them then we can start putting prices for them Something would be really useful that I've talked to people about doing is committing to buy a number of impact certificates at a Particular moment in time So if a large funder says we will buy five million dollars of impact certificates on this field Every year then there's a very strong Encouragement for participants all over the world to do a lot of work to create those impacts certificates and then compete for that for that funding at the end We can do this for decarbonization and so on and one of the good news is that carbon credits and Renewable energy certificates and a lot of other financializing instruments are exactly the kind of impact certificates That we're talking about those are just a special case Example in a specific field, but what we're talking about is something much more general something much more composable and something much more legible By the capital structures that exist today Like we basically want to come up with something that we can just hook into the capital structures and make them all work Cool, so with that last plug for sorry this get going it ought to correct it. That's so annoying How rude It's the most the most D gen thing ever instead of turn into the most region thing ever Yes, exactly if it only were that easy right Cool, so with that. Thank you very much Sorry for going a bit over over time and I apologize for having a new Mac book and filling us all with problems. Thank you