NFTs and ZK-proofs as Tools for Doing and Funding Good Science

Transcript

is shared in coded knowledge one other thing we think we know is that the incentives for producing this knowledge are abysmal so here's the thomas jefferson quote of all the things nature has made knowledge is the least suited to being exclusive property because the moment it's divulged it forces itself into the possession of everyone and the receiver cannot dispossess himself of it you can see explicit economic treatments in uh richard nelson 1959 kin arrow 1962 uh because of arrow's excellent treatment of knowledge as a public goods problem this is sometimes called the aerovian problem so this is thing two we think we know thing three is that incentives matter that's often called the core idea of economics is that you know people respond to

incentives well you stack these three things up and you get something pretty ugly uh we'll call it the try out of stagnation here right wealth comes exclusively from sharing encoded knowledge the incentives for sharing and coded knowledge are terrible and incentives matter good um well that seems to tell us the story that we're quite poor right that we're very very far away from the optimum if you took away any one of these things we might be okay but all three together are pretty bad well so positioning off of that uh you know transitioning to nfts might feel a little weird and grandiose but at least maybe i can convince you it's not so weird and maybe i don't have to convince people in this room but uh the wider world might still

think it's weird i'll at least say that you know i'm we've been beating the drum for a little while so my original uh argument on this was uh in this 2018 essay on nfts for science and the pitch was pretty simple historically great knowledge is valued like great art you can buy a page of darwin for a hundred thousand dollars um or only if you're willing to pay hundred thousand dollars i should say but today's darwin isn't hand scrawling notes on the beagle today's darwin is typing in a version latex file like the one i'm presenting to you on uh which means you need some sort of digital scarcity you're losing some sort of manuscript value and because science funding is so precious you shouldn't lose any of it so that's one reason you should do nfts for science

but there are some other interesting use cases this one i haven't seen but i think many of you have probably seen nfts as these kind of like permissioning devices for like things related to social tokens where you'll get into a discord and so on there's a little bit of this tradition in science as well you may know this quote if you speak ancient greek uh let no one ignorant of geometry enter here supposedly above plato's academy that one is actually really hanging over a door in the santa fe institute so there's some notion of an emergent commissioning that could be created with nfts that might be an interesting thing to explore um nft's prizes is one thing i'm excited about and if we have time at the end i'll try to come back and get to that but prizes themselves are actually valued by the market so this is

feynman's nobel prize which sold for almost a million dollars a couple years ago and so you start to get maybe the potential for a prize to be awarded in exchange for an nft and value to be created on both sides which could be a little interesting um so that's ways you could fund science but i think i promised in the title that i was going to talk about good science and you know you give funding you worry you know like new sorts of funding that goes outside of the bounds of conventional science you worry that you're going to take a shortcut um on snake oil pass basically right like so we have funding is bad in science but also the system kind of works right people we know we deny funding for good ideas

but we don't let a lot of terrible stuff slip through for long but there's a snake oil pass kind of popular way to do science which can be a little risky and so to talk about that danger and that risk a little bit it might help to sort of transition into an example so we can we can get into zk proofs without having to look at uh you know the the crazy moon math so let me tell you about the story of seth roberts so seth roberts was a berkeley psychologist he wrote this paper in behavioral and brain scientists or sciences on self-experimentation as a source of new ideas he exhaustively chronicled all these self-experiments he did some work some didn't on improving his sleep his mood his health his weight etc he found one thing that worked pretty

well weirdly enough he found it here in paris it was a it was a trip to paris and he found that drinking new sugary drinks reduced his appetite and so this connected with some of the like research he had done on rats and rat appetite and setpoint fat regulation and some things that honestly i don't know anything about but he wrote a paper on it and uh the idea was that new and unfamiliar calories reduce your appetite so it's like okay well how do you keep calories from ever being familiar you maybe make them flavorless so you can never get adapted to them this was an interesting theory it garnered a lot of interest in support you can see the new york times cbc less wrong a lot of people thought it was interesting it wasn't really to the level of science it was self-experimentation but it was

promising and you can see aaron schwartz who i assume you know many people know in here you can see this beautiful like succession of blog posts on his website raw thought where he's like i'm a pretty skeptical guy but i gotta admit this is intriguing enough to try then three months later he's like this is crazy this is the easiest thing i've ever done i lost a i lost 20 pounds blah blah and then next week he writes a future without fat okay we've solved as soon as the clinical trials for this thing roll out we're going to realize we've all just solved this thing what's the world going to look like like this ridiculous level of uh of excitement and optimism that i think is is characteristic of how how great he was but um the clinical trials didn't happen right like you can speculate about

reasons i have my own one thing is a flavorless calorie is just hideously unpatentable right there's so many ways you can uh you can literally hold your nose and eat evidently and so this never really became a finding we don't know if it works i mean i'm telling you it's compelling the evidence seems interesting it's anic data but it has never been validated via an rct well so um we did we have started to bridge this gap we ran a crowdsource kind of web 3 style study where the participants were awarded nfts and we had some kind of web 3 legends uh like zuko from zcash uh help out a little bit and publicize it and the results look pretty promising so um i won't get into this too much now but you're

welcome to ping me if you want to discuss it this is an nft of the results it looks pretty good again i would not say it's a scientific finding but we sold this nft for 24 000 that money is going to fund a replication which should start running here in the next couple weeks and uh that hopefully will be able to validate whether this is a full-on scientific finding that people have been sort of clamoring for an rct on for 15 years or debunk it a pseudoscience which we should do as well because of course you know this is science we're doing here we're not trying to just raise the banner on this weird theory it might be uh it might be wrong um i i mentioned all this in the last slide results look pretty good not satisfactory evidence we're running the

replication how to austin griffith for uh helping with the nft distribution and burner wallets um but so i just told you about this data it's really hard to convince people to not only drink olive oil to test some old theory but to also be willing to let you share their health data with everybody and share an open data set especially if you're just some guy on the internet like i was so there was this natural problem i mean i say my data says this but how would you know well zero knowledge proofs help us here and uh it goes just being it goes beyond even the sort of like hipaa death uh health permissioning sort of stuff this is endemic to science even when you have the rights to share this is a really really common occurrence right so you see a paper that

says data available on request and then they'll just never let you take a look at it um so a lot of times you have to trust you can't really validate this stuff now sometimes trusting isn't a problem you can see i promised here on june 19th i was going to include this slide and you just saw i did so clearly i'm like a very trustworthy individual but you know other scientists not so sure so we do these uh these zero knowledge proofs and we ran this uh with collaboration with elio and we wrote it in the leo language and and basically we were able to provide these two statisticians um a proving key along with um the proof that the slope of the line was what we said it was the slope of treat on uh hunger which is to say the result that we were describing in

that earlier nft and this is a pretty cool thing because it allows you to validate the claims and produce a proven key that sort of like preserves some of the structure of the data without having to reveal it because the truth is like as much as we want people to share data data is extremely expensive right people will pay a hundred thousand dollars for a data set and they don't want to share it with their first result they want to share it the way when they often do share it the way like a lion shares a carcass or something where like when they're good and done with it and they found all the good results you know you're welcome to step in but before that maybe not so so it's nice to be able to validate this stuff we published a nice piece on this well i

mean i think it's nice um and here's here's a statistician harry crane who collaborated saying they were able to independently validate the test of seth robert's appetite theory it's a big step in open science um it's kind of a cool thing um and so that's that's a couple of things that's that's nft tech that's zero knowledge proof tech there's some other cool projects that are going on in the space um expand on this so v to dao which i i think i need to lose that space in there um this has nfts as patent holders so like the nft can actually hold the patent itself which is a really cool thing and uh i think paul published on this really early

too which is uh cool we have ants review um and bianca is in the audience with us here very cool uh incentives for peer review which is a sort of nightmare problem in academia and an amazing thing to be solved um experiment.com they're doing nfts for scientific proposals which is a really cool thing um so the proposal itself gets nft and people can sort of own it as a patch um ops inches doing data wallets and then research hub has uh structured erc20 incentives for publication so all this stuff is pretty cool i think there's a lot of exciting tech to do good science with um one of the things that i'm not going to get into oh i should also mention nft of

some nobel prize winning research sold in may this was uh this was out of berkeley but that was a that was a really cool thing sold for fifty thousand dollars um one of the things i'm really excited about that i won't get into right now because of time constraints and complexity but i would love to talk to anybody about is we're um launching a system of self-funding prizes and this uses some interesting web3 tech notably smart contract commitment in addition to some of these nft texts we we talked about earlier but i want to close by highlighting something that i think would be easy to sort of forget about that's maybe the big deep secret advantage for doing science and web three and that's ultimately the culture right uh we're lucky to have people

heading up some really cool organizations who are deeply committed and interested in science so obviously the most prominent figure in ethereum vitalik you know he publishes stuff in economics he writes on cryptography there's a lot of value on cryptography computer science incentive design mechanism design in the space and that's a really cool and and treasured kind of thing so i think all these things combined it makes the uh the frontier of doing science especially with these web 3 tools a really promising thing we're seeing increasing interest i mentioned that prize we have i'm collaborating with researchers at santa fe institute columbia harvard a few others are involved in that so and these are these are researchers in the institutions

so they're they're coming around and getting excited about this even though you know when you've talked to them about it in may they were just wanting to ask about people um they're seeing so anyway um thank you all very much for coming out and hopefully this will be uh let's see desai summer turns out there's a question is there any uh live experimentation going on with zero knowledge proofs and

nfts that you can point us to uh can you see more the just the broad application is actually super interesting and it could be anything from you know people don't want to know that they own this piece of art to you know real estate to even the science stuff just the idea that nfts could transfer ownership could allocate economic benefits but you don't necessarily need to know specifics unless it's open to you by the owner so i mean we did do it we did do the experiment and actually didn't do that we sold the nft i didn't mention this in the slide but we sold the nft of the

independent data analysis that was done cpa approved that sold for i think 5 000 or something but it's i promise it's the first time the market's ever valued an independent data analysis so it's kind of a cool thing okay ongoing experiments i don't know but i will say elio has the code up for linear regression so you can do it um and i i think there's a lot of interest my suspicion is you'll be hearing a lot more about this stuff but i don't know if they have been at once okay if there's any links or anything that you can share if there's any links or anything that you can share later or privately or that'll be great oh here's mine oh this mic works too okay amazing

excuse both hi thank you um great presentation my question would be um what do you feel like is the most useful function of nft and funding science so you've said this nft was sold for quite a lot of money 24 000 what does it actually represent like it could also represent an ip like with retail what other ways do you see nfgs could be used and what's the most useful yeah it's hard to answer what's the most useful but i mean certainly holding real ip is that's an interesting use case an actual patent right um but i i do think it's interesting and in particular the way we've gone has been to focus on um alt ip so there's no

there's no actual legal right conferred so when we sold it it's just you own the concept of the result in the same way that when you own a page of darwin you don't own the theory of evolution you own something connected to it the theory of collectibles if you'll let me nerd out a little bit says that what's going on is that there is a biographical indexicality that gets created when people own something and collect it i know that's like a 10 word but this is a science talk so you should be able to be cool with it basically we like objects we like things that connect to important periods of our lives to important periods of history and so on right the sweater our grandma gave us the origin of the theory of evolution and so on right so if nfts are capturing this sort of like

connection in the same way you know the connection to the artist's original signature or the uh emergence of this idea then we think those those things parallel pretty well and you can get this interesting kind of like manuscript value where you own something that is increasing in value at least personal value as the idea takes hold so we hope that the person who owns this nft will come to treasure it more and more if it so happens that this theory is validated right i mean if it's this world-changing theory like sports that said it was i think it's really cool to hold this nft right so that gives you something like a flavor of the patent structure in terms of being able to reward early basic shared research without the government and forest exclusivity so somebody can start an extra light olive

oil company tomorrow and none of us have any claims against them right yes where do you see it be useful i'll repeat for the recording how do you see the change of um how do you see the government enforced um exclusivity rights and so on changing in the future thanks to web3 and crypto and do you think there are opportunities to actually improve science through those mechanisms yeah that's a that's a great and big question i'll try to do it justice i mean patents are not that old and people have complained about them since the very beginning right and so certainly there are people are well aware of the limitations of the system

i think understanding the reasons they exist and the problem they're trying to solve which is what i tried to get at at the beginning and i know you're well aware of of course um this idea of ensuring that shared high fixed cost knowledge is rewarded and preserved over time at least through attribution but hopefully through some bare minimum insofar as that solves the problem that patents are trying to solve i would expect you know the role of patents to reduce right like because it's a it's a costly you know centralized very imperfect sort of structure right will it always have its uses i mean i can't say i i certainly would expect in the near term patents or patents are here to stay and you know i think there's a lot of use for those projects but i think it would be really cool to do

something that served what the the purpose of patents without having to be this centralized exclusive thing you can kind of get the dream of the open internet where you're able to share this thing openly everybody can use it immediately but you also get rewarded right similarly with patents yeah thanks yes uh i was wondering in the context of uh what you did with this uh zk proof like you can prove that the computation was done correctly but what about proving that the data is the correct data like is there anything about you know maybe like having um the participants

in the in the study somehow have you know some input that validates that you know the data that you actually provide the proof for is the correct one or whatever some kind of thing of the along these lines i think that's that's a really interesting idea um and i haven't thought of that but the idea that the individuals who participate in the study would in some sense validate it i mean it does just push the trust problem further upstream but it does you know arguably decentralize it which could be kind of cool so your point about zk proves absolutely true i will say that um and just to be clear his point is that um yes you can tell that the data says what i says it or what i what i say it does but i you can't tell whether the data is real or not the

the work that i've seen done on what actual like p hacking and like falsification of research is is it's very rare for scientists to just outright make up a data set it happens but it's extremely rare what's more common is this sort of like subtle i'm gonna drop this outlier i'm gonna modify this when the referee tells me to do something right like it's this more subtle psychological process that i think we can all sort of intuitively recognize when we're tempted to do something we know is wrong you kind of like massage it and so sharing a totally fraudulent data set you'd certainly see it and zk proofs can't catch you on that but i think they're much more common and here i'm citing work from uh from andrew gellman who runs statsblog uh the the idea that there's more it's closer to a garden of

forking paths where as you deal with the data you start to make decisions unconsciously or subconsciously or semi-consciously that push you toward your results now i think zk proofs can help with that especially as the process is iterated right if somebody says like well what about this what about this you can't go back and change your data after the fact is that does that answer your question yeah okay great good all right thank you [Applause] you