No Incentive
Speakers: David Schwartz
Transcript By: Bryan Bishop
Tags: Incentives
Category: Conference
The best incentive is no incentive
https://twitter.com/kanzure/status/1230997662339436544
Introduction
This is the last session of the conference. Time flies when you’re having fun. We’re going to be starting with David Schwartz who is one of the co-creators of Ripple and he is going to be talking about a controversial topic that “the best incentive is no incentive”.
I am David Schwartz as you just heard I am one of the co-creators of Ripple. Thanks to Dan Boneh for inviting me to be here. I wanted to give this talk 7 years ago but there wouldn’t have been a place to give this talk. It’s nice that there’s a shared understanding now and I can explain the decisions we made back in 2012-2013.
I feel like I have to give a trigger warning here. If there are any bitcoin maximalists in the room, some of this material might be triggering for them.
The problem
Why do we need incentives in blockchain? Maybe everyone can run software and everything will just work. Public blockchains have characteristics like, the system state is public and every honest participant in the system knows which transactions are valid or not and they know the rules for determining what a transaction does. But geeze, isn’t that enough? If we can all agree on what transactions are valid and what they do, what do we need incentives for?
It’s important to keep in mind that we don’t care what dishonest participants can do to themselves or each other. In bitcoin, I can modify my software to say I have a trillion bitcoin but it doesn’t impact anyone else. But what about the people that follow rules and want to get the benefits of the system?
The problem is double spending. I think we all know this. Alice needs to know that payments will be accepted and we need eventual agreement for the system to be useful. If you can send coins, then you can send the same coins to some other place. The problem is that there’s more than one way to make valid forward progress. That’s why we need incentives.
If there are objectively better ways to make forward progress- like by executing no transactions at all, or by executing a valid transaction, all participants will agree that executing the transaction is better than not executing it. We don’t need incentives: the honest participants will just agree on the better transaction.
The problem is when there’s two equally good ways to make forward progress. Honest participants don’t really care to choose between the two good options. We just want to not have disagreement; we don’t care who someone pays. If she submits conflicting transactions, then she gets whatever random outcome, we just need to agree on the outcome. When we have two equally good ways to make forward progress, how do the honest stakeholders eventually agree on this?
That’s the actual problem.
Stakeholders
I want to talk about stakeholders: they are people who get value from the system. They have a problem that the system solves, and they want the system to function. If the system is ebay, the stakeholders are people who want to buy things, people who want to sell things, and ebay. They all want ebay to function reliably and smoothly but their interests are not perfectly aligned. Buyers and sellers probably want low transaction fees, and ebay probably wants high transaction fees. Stakeholders have at least partially aligned incentives, but not necessarily identical incentives.
Natural stakeholders
Natural stakeholders are people who the system solves problems for. Miners are not natural stakeholders for miners. Bitcoin was not made for miners to turn electricity into money. Bitcoin was built to solve the problem of having a means of exchange or having a store of value, and those are the natural stakeholders that the system solves a pre-existing problem solved and they are willing to pay to get the benefits.
Forced stakeholders
Forced stakeholders are the people you require to make the system work. This is the push for decentralization is to eliminate the forced stakeholders, the people who extract value from the system that the system hasn’t removed yet. One of the core pitches of decentralization is to get rid of some of the forced stakeholders. The forced stakeholders impose rules that the natural stakeholders that they would prefer to not follow. This is friction that the natural stakeholders whose problems the systems solve, they want low fees, and forced stakeholders want high fees. If the system is too expensive, then the system won’t solve any problem, so the incentives are somewhat aligned but not completely aligned. Natural stakeholders want the system to solve their problems cheaply. Miners are forced stakeholders in bitcoin. You want low transaction fees, high transaction fees don’t benefit you except that they make the system work. Natural stakeholders prefer as little value go to them as possible. Natural stakeholders always want the middlemen to go away. One of the big promises of blockchain systems was the reduction of middlemen. The pitch was be your own bank. The middlemen either don’t exist or their role is very minimal, that was the pitch.
Operational decentralization
Operational decentralization is assurance of inherent fairness. Operational decentralization is assurance the system operates fairly. It’s short-term assurance of fairness. It doesn’t exclude long-term concerns. Operational decentralization with bitcoin means that if I submit a bitcoin transaction right now and it pays fees and it’s all valid, can I have reasonable assurance that it will execute and the system won’t single me out? It’s a form of short-term censorship, it’s independent of the question of whether bitcoin will be the same in 10 years or whatever, will the rules change or something? That’s not part of operational decentralization, which is only about the short-term guarantees that your transaction will execute according to the rules without bias.
Governance
Governance is what stops the rules from changing in bitcoin. Governance is how you stop technological obsolescence. Avoids chaos when the system rules need to be changed. Natural stakeholders don’t want their system to become technologically obsolete. What we mean is what the stakeholders can realistically expect… most people will argue there will only ever be 21m bitcoin in existence because if it was 22m bitcoin it wouldn’t be bitcoin anymore. If I’m a natural stakeholder and I have bitcoin, and the miners agree to raise the block reward, and there’s more bitcoin, and my old system is defunct and my new bitcoin is on a new fork or something- from my point of view as a natural stakeholder if I am relying on there never being more than 21m bitcoin then that requirement was violated. If the system stops working and fails, if I’m a stakeholder that wants the systme to continue to work, I wouldn’t be happy with a system that has 21m bitcoin and isn’t useful. What is the stakeholder actually going to get out of the system, instead of what definitional tricks we can play.
Proof-of-work
Proof-of-work was the solution proposed in 2009. The biggest problem is that honest participants have to actually pay as much as any attacker is willing to pay. The natural stakeholders, the people paying the miners to mine and the people whose money is coming from outside the bitcoin ecosystem and going to build ASICs and paying electricity, they have to pay on a continuous basis more money than any attacker is willing to pay to execute double spend. Imagine that I had to constantly pay $300 for my TV because an attacker is willing to spend up to $300 for a TV then I haven’t gained any benefit from having the TV because I have to constantly pay its value just to keep it. Bitcoin pays millions of dollars per day just to move money. This money goes from natural stakeholders to artificial stakeholders. It’s no different from other middleman. You could argue miners are stakeholders that want to turn electricity into money, but if you don’t care about that then from your point of view that’s friction and miners want high fees and people transacting want low fees.
Double spending risk
It also creates the risk of a double spending attack. If the token doesn’t have value, you can do a double spend attack. It’s not clear how to recover from this. If you don’t change anything, then you’re still just as vulnerable, or if the token value drops then you’re even mor evulnerable. If you didn’t have enough honest miners before, and you do an ASIC hard-fork, then you turned their ASICs into space heaters a little faster than they would otherwise… good luck encouraging more honest ifmining you do that…. This destroys your ability to use the system and to rely on the system.
Advantages of Proof-of-Work
From what I just said, it sounds like bitcoin is awful and terrible but no it works just fine. You can have a lot of objectively bad characteristics and the system just works fine. This is a profound need that is being satisfied: it can have all these technical problems I’ve pointed out, but people still find it useful and they believe people will find it useful in the future at least.
Currency distribution
Proof-of-work can do the initial distribution of a currency. If I give away my phone to someone and everyone is going to want the phone. There has to be something expensive to stop people from getting it. Proof-of-work can do the initial distribution of the currency, it’s hard to imagine doing this with another scheme that didn’t have high cost.
Miners
Miners take money out of the system and take a profit, they are classic intermediaries but not necessarily a bad thing. They work fine. Bitcoin still works. It’s still the grandaddy of them all… Proof-of-work relies on native value, though. You have to pay real money to the miners. The security is tied to the value of the currency. Ethereum hosts ERC20 tokens that aren’t ether, and so proof-of-work wouldn’t work. It’s hard to imagine how to meet the security requirements of ERC20 tokens. Suppose there’s a token pegged to the dollar, and the ETH price falls. How do you get the system to be secure? How do you secure the movement of ERC20 token? It’s not clear that the system can capture the value of the other systems riding on. Maybe moving millions of dollars of value has a near zero cost. The cost of moving things has come down to zero. It might be useful for values, but proof-of-work doesn’t work in that situation so it’s not a good choice to build on and it has a race to the bottom.
Race to the bottom
A miner can try to get all the DeFi transactions in the ecosystem. They can game the system and not allow people to cancel orders or whatever. He is going to make more money than the miners that don’t do that. Mining is the race to the bottom and you have to get more money out of every hash you do. There’s no incentive to improve the network other than a diffuse incentive that if the whole system is better than the ecosystem is healthy. We should presume miners are selfish. I don’t want you to mine because you’re pushing up the difficulty, you’re indifferent to the network quality because you’re forced to be this way.
Does not operationally decentralize
All the people who mine tend to look similar. Cheap power, a lot of equipment. All the miners must be able to operate cheaply. People with these abilities tend to be similar in many ways. Miners choose how the system makes forward progress.
Staking and slashing
Staking and slashing didn’t exist 7 years ago. You lock up a volatile asset and you lose it if you mess up. The returns have to be high to justify this. If you had to lock up a bunch of bitcoin, well these assets are very volatile and you should be paid a good amount of money to take this risk. You lose the amount if you break system rules. So there’s risk. The returns would have to be significant to justify the risk. The security is tied to the asset value, so it’s not good for systems where most of the value isn’t in the native token.
Staking rewards are taxable. Everyone is paying taxes on this just to stay put. Stakers are forced stakeholders solving the natural stakeholders problem. If the pitch is ebay without ebay, no intermedaries, then there shouldn’t be forced stakeholders.
Incentives are expensive
Incentives tax the natural stakeholders. In some schemes, attackers are paying more than attackers are willing to pay. Artificial stakeholders want high fees, they want as much friction in the system as possible and still have it attactive to the natural stakeholders. They will settle for low fees, but tha’ts not the promise right?
Exploiting natural alignment
The alternative is to exploit natural alignment. Natural stakeholders are already aligned. They all want the operational decentralization, they want it cheap, etc. They are all aligned on wanting to reach agreements on equally good ways to make forward progress. None of them care which of the good ways are chosen, just that they are aligned. Why can’t they solve this problem for themselves?
The answer is that if you bring in incentives ,you break that. The only reason you need a high incentive is because mining is expensive. It’s a self-referential circle. Natural stakeholders want that cycle broken. Cheap systems can exploit natural incentives.
You need something scarce
You need something scarce ohterwise you have a sybil attack. You have two equally good ways to make forward progress and a sybil attack can make sure you never reach consensus. The XRP ledger uses stakeholder-chosen scarcity so you just choose anything scarce. It minimizes operational power, no dictator of the moment, it sets power to objective rules to make rules for which transactions get included in each block. We don’t consider invalid transactions. The only way to reject a valid tansaction is to agree to include it in the next round which comes in the next 5 seconds. Stop listening to nodes that are part of the sybil attack. Fake stakeholders get ejected. There’s no advantage to having bad actors in the system. If someone is mining and gaming a decentralized exchange, we would want to stop them if we could but we can’t. We want the network to be good, you want the cost to go down, no artificial incentives. There’s no double spending, there’s no attack, you can only choose between ways to make forward progress so why would you bother attacking the system? There’s no incentive to attack the system. Stakeholders get what they want- minimize drama, risk and cost.
Natural versus artificial incentives
Natural incentives are better than artificial incentives. Gives you low fees, fast, censorship resistance. We built a decentralized exchange, we have off-ledger scaling. Come to xrpl.org and see what we built and Ripple is always hiring.
Sponsorship: These transcripts are sponsored by Blockchain Commons.
Disclaimer: These are unpaid transcriptions, performed in real-time and in-person during the actual source presentation. Due to personal time constraints they are usually not reviewed against the source material once published. Errors are possible. If the original author/speaker or anyone else finds errors of substance, please email me at kanzure@gmail.com for corrections or contribute online via github/git. I sometimes add annotations to the transcription text. These will always be denoted by a standard editor’s note in parenthesis brackets ((like this)), or in a numbered footnote. I welcome feedback and discussion of these as well.
Tweet: Transcript: “The best incentive is no incentive” https://diyhpl.us/wiki/transcripts/stanford-blockchain-conference/2020/no-incentive/ @ripple @CBRStanford #SBC20