The ACM Special Interest Group on Artificial Intelligence recently started publishing a quarterly newsletter called “AI Matters”. We submitted this article to describe some topics that our Bay Area SIGAI Chapter is currently exploring:
Cryptocurrencies, Smart Contracts, and Artificial Intelligence
by Steve Omohundro
Recent developments in “cryptocurrencies” and “smart contracts” are creating new opportunities for applying AI techniques. These economic technologies would benefit from greater real world knowledge and reasoning as they become integrated with everyday commerce. And cryptocurrencies and smart contracts may provide an infrastructure for ensuring that AI systems follow specified legal and safety regulations as they become more integrated into human society.
Humans have attempted to create secret codes since the invention of writing but modern mathematical cryptography has been developed over the past 50 years. Public key cryptography was introduced in 1976 and the first cryptocurrencies were attempted in the early 1980’s (Griffith, 2014). Digital signatures can guarantee that a transfer of digital cash was authorized by its owner, but cannot prevent the owner from spending cash twice.
Several solutions for the double-spending problem have been proposed but the 2008 posting of the “Bitcoin” paper (Nakamoto, 2008) by pseudonymous author(s) Satoshi Nakamoto described the first system which has become widely used. Bitcoin solves the double-spending problem by creating a global decentralized ledger called the “blockchain” which records all transactions and prevents bitcoins from being spent twice. The current value of all bitcoins now exceeds 5 billion U.S. dollars. The bitcoin source code is freely available and 507 other cryptographic coins called “altcoins” have been created on the same principles (Coin Market Cap, 2014). A flurry of bitcoin and altcoin based businesses have also been created ranging from coin miners and exchanges to coffee houses and flower stores (Coin Desk, 2014).
The bitcoin blockchain establishes a decentralized consensus about the order of transactions among a large number of agents who don’t know or trust one another. If this were attempted with a simple consensus mechanism like voting, then dishonest agents could create many copies of themselves to manipulate the vote. Bitcoin prevents this kind of “Sybil” attack by requiring the agents involved in consensus formation to perform expensive computational work called “bitcoin mining”. To manipulate the consensus, a dishonest bitcoin miner would need to perform as much computational work as all the other miners combined. The miners are willing to perform the consensus function because they are rewarded in bitcoins for their efforts.
Bitcoin supports some forms of transaction that go beyond simple transfers of coins from one party to another. For example, it is possible to implement “multi-signature” transactions in which two out of three (or, in general, m out of n) participants must validate a transfer (Bitcoin Wiki, 2014). But bitcoin’s facility for defining contracts is limited.
Contract law has been critical to the formation of sophisticated human societies. Digital, self-enforcing “smart contracts” were proposed by Nick Szabo in 1993 (Szabo, 1997) but the economic and communications infrastructure at that time weren’t adequate to support them.
With the success of Bitcoin, several groups have proposed successor “Bitcoin 2.0” designs that incorporate more sophisticated forms of smart contracts. The most developed of these systems is called “Ethereum” (Ethereum, 2014). It has a blockchain similar to bitcoin’s but has a Turing complete contracting language which executes on the blockchain. This capacity allows complex contracts to be created and automatically enforced.
Ethereum’s facility for complex contracts allows financial exchanges, insurance contracts, financial derivatives, and many other kinds of transactions to be precisely defined and executed. Digital services like the rental of digital storage, computational power, and communication bandwidth are also easy to implement. There are also proposals for extending contracts to include information and interactions in the physical world. These include reputation management, “smart property” ownership records for real estate and vehicles, earthquake or weather insurance, and automated room rental (Buterin, 2014).
Modern corporations are defined by a set of contracts with investors, management, employees, customers, and suppliers. If these contracts are automated, then “Distributed Autonomous Organizations” (DAOs, sometimes also called “Decentralized”) become possible. These entities might buy and sell things, make decisions, and hire and fire contractors without human management. It is also possible to create human-run organizations which make decisions by voting on the blockchain. Adam Levine has proposed “Self-bootstrapped” organizations which issue cryptoequities to investors based on a mission statement and then create themselves using contractors guided by decentralized blockchain voting (Levine, 2014).
The ultimate expression of these ideas is the “Distributed Autonomous Society” (DAS, sometimes also called “Decentralized”)(D’Onofrio, 2014). Many of the current functions of government might be implemented more reliably and cheaply using smart contracts. For example, BitCongress (BitCongress, 2014) is a blockchain based voting system. Other proposals suggest blockchain implementations of taxation, the Federal Reserve, intellectual property, universal basic income, real estate records, etc. (Prisco, 2014).
Interaction with Artificial Intelligence
The social issues raised by these emerging technologies are enormous. The “Internet of Things” is predicted by Gartner to include 26 billion devices and to result in $1.9 trillion in economic value-add by 2020 (Gartner, 2013). Many are predicting the need for a corresponding “Internet of Money” to manage the transactions between these devices. This is likely to involve cryptocurrencies and smart contracts which will then come to interact with every area of our lives.
Simple and easy to write contracts appear to be sufficient for many entirely digital transactions. But as these systems start to interact with the physical world, there is likely to be a need for greater intelligence and real world knowledge in making decisions. AI systems will be needed to translate information from a wide variety of sensors into precise terms that smart contracts can act upon. In the other direction, contracts that lead to physical actions (such as delivery of items) will need to interface with human and robotic agents. For example, farmers might want insurance contracts against harmful weather conditions and a smart contract would need to determine when the payout clause is triggered.
The legal codes of many countries have become quite complex. Several AI projects are trying to create formal digital versions of legal codes (CodeX, 2014). These systems will eventually be used to resolve legal issues and perhaps even act as arbitrators or judges. Sophisticated AI systems with knowledge of the legal system will be used to help craft and simplify new legislation.
Cryptocurrencies and smart contracts may also play a role ensuring that AI systems are beneficial for human society. For example, self-driving cars need to follow the rules of the road, autonomous business creation needs to follow securities laws, and autonomous markets need to levy taxes appropriate for transactions’ jurisdictions. Over the next few decades we will need to extend many human laws and ethical norms to automated robotic and intelligent systems (Omohundro, 2014). Cryptocurrencies are a natural way to implement the economic transactions of these systems. Smart contracts are a natural way to impose legal and safety constraints on their behaviors. But many new insights and innovative ideas are needed!
Bitcoin Wiki. (2014, November 9). Contracts. Retrieved from the Bitcoin website:
BitCongress. (2014, November 9). BitCongress – Blockchain Based Voting System.
Retrieved from the Ethereum website: https://forum.ethereum.org/discussion/110/bitcongress-blockchain-based-voting-system
Buterin, V. (2013). Ethereum White Paper. Retrieved from the Ethereum website:
Coindesk. (2014, November 7). What Can You Buy with Bitcoin?. Retrieved from the
CoinDesk website: http://www.coindesk.com/information/what-can-you-buy-with-bitcoins/
CodeX. (2014, November 9). CodeX: The Stanford Center for Legal Informatics.
Retrieved from the Stanford Law School website: https://www.law.stanford.edu/organizations/programs-and-centers/codex-the-stanford-center-for-legal-informatics/projects
Coin Market Cap. (2014, November 9). Crypto-Currency Market Capitalization.
Retrieved from the Coin Market Cap website: http://coinmarketcap.com/
D’Onofrio, A. (2014, November 7). Distributed Autonomous Society. Retrieved from the
Quora website: http://distributed-autonomous-society.quora.com/
Ethereum. (2014). Retrieved from the Ethereum website: https://www.ethereum.org/
Gartner. (2013). Forecast: The Internet of Things, Worldwide, 2013. Retrieved from the
Gartner website: https://www.gartner.com/doc/2625419/forecast-internet-things-worldwide-
Griffith, K. (2014). A Quick History of Cryptocurrencies BBTC – Before Bitcoin. Retrieved
from the Bitcoin Magazine website: http://bitcoinmagazine.com/12241/quick-history-cryptocurrencies-bbtc-bitcoin/
Levine, A. (2014). Application Specific, Autonomous, Self-Bootstrapping Consensus
Platforms. Retrieved from the Bitsharestalk website: https://bitsharestalk.org/index.php?topic=1854.0
Nakamoto, S. (2008). Bitcoin: A Peer-to-Peer Electronic Cash System. Retrieved from
Bitcoin website: https://bitcoin.org/bitcoin.pdf
Omohundro, S. (2014). Autonomous Technology and the Greater Human Good. Journal
of Experimental and Theoretical Artificial Intelligence, 26(3). Retrieved from the
Taylor and Francis Online
Prisco, G. (2014). Bitcoin Governance 2.0: Let’s Block-chain them. Retrieved from the
Cryptocoinsnews website: https://www.cryptocoinsnews.com/bitcoin-governance-2-0-lets-block-chain/
Szabo, N. (1997). Formalizing and Securing Relationships on Public Networks. First
Monday, 2(9). Retrieved from the First Monday website: http://firstmonday.org/ojs/index.php/fm/article/view/548/469