Discover more from Yuxi’s Substack
Blockchains Require Dramatic Innovations to Prosper
Blockchains calls for killer apps
Deep learning has revolutionized computer vision and natural language processing (NLP), with applications like face recognition and ChatGPT
Reinforcement learning has breakthroughs like AlphaGo, but so so in business
Blockchain: Crypto? NFT? DeFi? GameFi? Seems not quite revolutionary yet.
The current practice is misaligned with the goal.
A codebase is controlled by a small group of people.
Vitalik Buterin is the King of Ethereum.
Blockchains are built on the Internet, which is controlled by authorities.
Blockchains are more or less, sooner or later regulated by some authorities.
The influential crypto exchanges (Binnance, Coinbase) are centralized.
A blockchain may encounter soft- and hard-forks.
Matthew effect: The rich get richer and the poor get poorer.
No cryptocurrency is stable yet.
The current codebases are far from mature.
“Smart contracts” are neither “smart” nor “contracts”.
basically “stored procedures”
e.g., All Smart Contracts Are Ambiguous, J. of Law & Innovation, 2019
e.g., A Survey of Smart Contract Formal Specification and Verification, ACM Comput. Surveys, Jan 2021
The current codebases are very vulnerable to attacks.
e.g., A Survey on Ethereum Systems Security: Vulnerabilities, Attacks, and Defenses, ACM Computing Surveys, June 2020
Tradeoffs are required due to the security-decentralization-scalability trilemma.
Blockchains are run by humans, with all potential good and evil, i.e., not ideal.
Pragmatically, blockchains need to co-exist with other forms of institutions, like firms, markets, economies and nation states, for a long time, or FOREVER.
Blockchains are innovations, w.r.t.
An information and computation technology
“world computer”, Internet 2.0, Web3
An organization and governance technology
competing with firms, markets, economies
e.g., Bitcoin’s academic pedigree, CACM, Dec 2017
e.g., Some Simple Economics of the Blockchain, CACM, July 2020
the cost of verification and the cost of networking
e.g., Blockchains and the economic institutions of capitalism, Journal of Institutional Economics, 2017
Cross-disciplinary by nature
Computer science, economics, behavioral science, optimization, etc.
e.g., Foundations of Cryptoeconomic Systems, 2020
e.g., a MOOC, Zero Knowledge Proofs (2023)
Distributed computing; consensus protocols.
e.g. A Survey on Consensus Mechanisms and Mining Strategy Management in Blockchain Networks, IEEE Access, Jan 2019
a building block for all blockchains
a building block for all (?) “Blockchain 2.0” applications (e.g., DAOs)
Desired properties: decentralized, secure, scalable, immutable, trustworthy, open, transparent, persistent, resilient, interoperable
Too complex for theoretical analysis / modeling
usually too large theory-practice gap, due to too strong assumptions
Resort to computational approaches
Simulation, Reinforcement learning / AI
Gamification is relevant
A course on Coursera
A blockchain simulator?
e.g., Agent-based modeling in economics and finance: past, present, and future, Journal of Economic Literature, 2022
e.g., "Auction Learning as a Two Player Game": GANs (?) for Mechanism Design, ICLR 2022 Blog
Ongoing efforts (from the community)
Complex Adaptive Dynamics Computer Aided design (cadCAD)
e.g., Beyond markets and states- Polycentric governance of complex economic systems, American Economic Review, 2010 (Nobel Lecture)
e.g., Automating Ostrom for Effective DAO Management, 2019
The above was basically written at the end of 2022. Everyone knows what happened then. It is hard for a person in AI to resist the big wave.
There may be good opportunities at the intersection of AI and blockchains. One example is credit assignment, e.g., for open source software, which is indispensable for further progress of AI, yet may lack of a proper incentive mechanism. See the recent Future of Decentralization, AI, and Computing Summit. I plan to spend more time on this promising direction.