Against DeFi Oracles

Transcript

Oh moment okay I'm Tilly Kamal should be second time your secret to tell you I also Daniel do that why it's attract a

much of the colony cultural knowledge no Betsy wetsy come to Calvinism lon originally they said well you cool Department would you always complain okay

Antonio when you cannot scream you take reckless not honest keep a tidy profit right back to this amazing it's obviously not something we call inertia okay okay so much DoCoMo sister

[Music] I hope you can hear me all right and I'm sorry about here today I because a virus in Singapore but I'm really happy to present that in CC it's really nice I was there last year I want to thank the organizers for arranging the meeting and if you cannot see sight valina you can also find them on each other / - RIT great okay let's stay and I couldn't be stopped really we have s so this isn't we addressed its to fish - from the South

of France and it has lots of things that Tina and I think miss - is also a lot of things but I wanna give an introduction to what we is Provost incentive group and I want to show you like a few things that we've been working on in the last applauds so the robust incentives groups is new research team at the theorem which is we focused on everything that has a strategic flavor and is pathetic I mean that we look at situations where you have a set of stakeholders they have to make decisions they make decisions that we optimize their own games the payoffs are affected by what people are also design so I'll give a few example

to be less general but you can already see that it's a it's very wide model and certainly whether we can use to study protocol security but also decentralized applications from a strategic point so what do we mean by incentives so we eye of the Ring we really mean that totally key we use is based on these three let's see academic field the first one is in theory which probably have about if you have work to be doing yet you still find players who have a certain set of moves actions and the moves I want to give you is to try and analyzing

and so the question well what should I do in a strategic situation six very descriptive acting me let me have meetings and design which is sometimes called inverse in theory which is you start from the end you say here's like an outcome that I had a situation like a social situation of social configuration that I like and let me try and design game so that players interacting with it and achieve that outcome on that and then there's the final strength which is relatively ninety-nine we care about designing these mechanisms we care about

finding good games that give us good outcomes we also care that the mechanisms are computationally tractable meaning that they can be achieved by agents naturally founded by founded we can request access agents are running algorithms and they don't have as much computational power as much storage as they want they're still trying to make the best decisions so this is sort of a general idea where we are locating a search by robust represent comes from the fact that [Music]

tomorrow is 2 and they're really complex and then just complete is because they contest but whatever context because they are the mishmash of many different products and this complexity just naturally compounds and so you can have very complex behavior emergent behaviors and you want to protocol to survive in any mobile service functions we spend the tax in certain conditions and to feel about robustness is actually very different in these we are doing what I can in vacuum game if you know game theory might not expect prisoner's

dilemma miss Kim's game they really abstract settings the initial test protocols there's a lot of things we don't have first which our players we multiple talkers to be robust against which is a games we also not be able to place at the same equations we don't like the system and it is probably not the case we also care that players are

meeting a larger because in space this is what we mean and finally one in my group I think very interesting focus of the ring is to try and develop outreach education also awareness in strategic questions exist and we try to formalize the rifles along three tracks first find the resources that allow us to understand [Music] [Music]

[Music] [Music] and to make sure that we device we also expect them to create gloves when we we

expected to look around and check if all the leaders behave nice inators we should look for visible and then in later phases we also expect so how to be in sense it work so it some kind of stage we try to make sure that we'll lose money well we and it's the very super just be multiple management what can we expect

this is what we want to try and I will tell you that to really understand okay well can we expect from protocol we can find our economic analysis in two different tracks first one has to do with let's a protocol stability meaning that we specify some desired a little behavior on something that poor and visually exhaust will behave degrees and then specify some big ones we say if you move east this is the article that is happening you will get back we well you see you get a team and we hope somehow that very well that we define the game with the defining is [Music]

different at the assembler worries to security of the chain and it is definitely less attacks we cannot at this meaning then as bad as possible for both we have a really nice weirdness lots of things in indicator and introduce first decide your protocol states you can think of

space to space where the space is defined by the parameters of the protocol and one instance is determined by these parameters so which parameters am I talking about you can say for instance the minimum deposit the reward schedule meaning what what we eat Veritas when we when we behave nicely the time to exit to get your deposit back once again also things like the issuance curve which defined on your deposits that you get or behaving in s the thing and featuring on the right is taken from the post I wrote

with Sasha on - research we can actually drag these parameters around and see how we want our validate sausage in so recently what we mean in processes very nowadays many different parameters one protocol out of space by just setting some of its business the first thing that I really want to help everyone to be stable expected and then this is happening that's what we called incentive combat

we miss you we say here well it's rational for the middle to actually to do that and so we have this very large space in this notion of stability from a paper written dive in here in Singapore which actually looks at many of the different facets okay so I said I will give you some examples instability the

most [Music] but what eating cereal sound with the net mining power it's a better response for me as a miner - sometimes no trees will go back and keep it to myself some time and again it's a nice chain and then waste everybody else's time by saying hey I have a bogus change so you have the flu so this is this is the little behavior but we actually have a more recent example you can see in this paper with

the timing of your release can increase the payoffs of the pedicles I can be strategic about what I'm doing and I reach from vertical stability to asset security so we respected already protocol space to a smaller space of processing the witness and forget that these protocols we in our economic space I really like this tickle fight has to be going this is where concentration - isn't quick and the key to understand this is to really understand that when

of course I care about maximizing but then we're still going to think about my attack is it imagine for me - its gradual process of becoming taking an attack rather than just keep mining honesty and getting will become reversed threading so we hear that we completely understand rational behavior of the religions if we just protocol behavior and so given that we would like them to stay with which other

parameter that we should set for when we always try to optimize something and with something would be the civil security and we know that there are trade-offs between increasing issuance of decreasing decreasing issuance and we know that is rate of swimming surfing industry impacts how we set we want to try and write them down but actually moving so far has been more focused on looking at protocol stability for each stripe in down magnifico specification [Music]

for we speak on rather we at a rope chains as control dynamical systems the ecosystem is very well research ethics and basically start from state to state I cannot take my state by chain these resurrections are retreating it's a cat-cat wrap of the two specs

which is really focused on that little behavior and insensate and what is cat can to really nice pneumatic framework which is focused on submitting miss context dependent we have stress this idea of [Music] [Music] [Music]

this is an exchange where are they what can we expect the future and you also

might not be able to simulate the dynamics of recovery from change and that's how it attacks how Thanks we do check precise [Music] and I want to finish with a few more things interesting [Music] [Music]

[Music] [Music] [Music] [Music]

[Music] thank you to do questions do two questions do two questions so based on your research hi and thank you for the

presentation based on your research what is the your your assumption regarding the annual rate of return for for valid data I know there's this common assumption that 20 million East would be staked but the staking ratio in network like this is as high as nearly 80 percent currently so in a scenario where you have you know large custodian exchanges offering those taking facility like the currently do for POS chains to the existing enormous customer base and we reach you know staking ratio of like that is higher

than 80% or 90% what would you what's your yeah what do you have any specific opinion on this and what's uh what's your projection as well for that [Music] [Music] [Music]

[Music] [Music] [Music]

okay yeah thank you yeah I mean it makes a lot of sense I think one additional question then is based on the mechanical design approach sort of - how you Muro that approach of token issuance and and the reward function in if - and it's related to the decentralization of the network so I know we want an incentive for high uptime we want to incentivize for as much as we can to be staked to guarantee the security of the network but have you taken into account or considered putting into your model incentivizing people from moving away

from cloud instances for instance I read this figure like today or yesterday that stated that 60% of who nodes operating on a theorem right now are actually based on AWS is that a concern to you and do you think that could be some consideration some consideration for how cutting those rules into the protocol saying you know you'd be you'd get higher reward from running a footnote from your home rather than you know just using some some instance I want to know like yeah what's your opinion on that and if any work has been undertaken to some extent I think you because the

protocols already quite complex you might want your protocol to be agnostic to how its brain so your encoding in your protocol some real-world rules that say oh I noticed that you're using like AWS or something you already kind of biasing your protocol to some extent actually you make me think of like this Vitalik based on credible mechanisms like you might not want to say obviously outcome they expect that you might not that protocol level to say well here is like however it was function and then possibly try to find out different mechanism on top to to incentivize let's say not using AWS there's one thing and so if you check the specs of each - there is some kind

of penalty for let's say correlated attacks or correlated bad behavior so when too many people are doing the wrong thing at the same time more of a share of the deposit is lost and one ID behind having this misfit one is also to say that if you have located all of your nodes on the same server and your server crashes you're going to lose proportionately more but your deposit and so we this this was put inside women who mechanism so that it doesn't sensitize people to run like it's a different thing look at these different versions so but not everything crashes at the same time in which case like we've been empty is much higher okay

yeah thanks thanks very much for the answers yeah thank you good question hi Dmitry here related to the last part I think it's a bit I feels a bit like similar to getting a sick resistance in the network and that you have like CloudFlare resistance or basis resistance into the network but regardless of that I know you touched it slightly but I just want to hear your opinion about incentives the security side of things first is incentives that are now running on the D upside on the D Phi space and stuff like that you see compound and stuff going through 7% interest rates are you factoring that in when you model I know it's very hard to

factor in it looks and feels a bit like an externality to to the protocol incentives is there any methodology you apply there like to I mean like these things are really hard to formalize I'm hoping simulations are better to to look at these questions but most specifically on this idea let's say I am a validator I manage some kind of portfolio of assets I decide to allocate my assets between staking or between lending and compound again I would refer you like to this paper by Tarun Chitra which is in

the slides but yeah I think one conclusion that Tarun has been talking about for some time is that the protocol should have some kind of awareness of what the chain is proposing in terms of returns and he said I don't think it's like a no reasonable idea but I but to link actually to the last to a previous question and it's it's kind of uncomfortable to but your protocol needs to depend on some sort of Oracle of is happening on the chain because again I think yeah it is really to have a protocol be as unbiased and as yeah almost like blind by design of what is happening but but this raises very

serious questions and the I think the paper by Tarun makes a really good point to say that these are actually things we should be looking at and they will they should probably impact somehow the design I'm not certain how yet I not true it's so easy to model as well but I really new hope but doing simulations with clarifying some of these points yeah okay [Applause]

you