Can AI solve centralization issues for blockchain projects?
There’s more hype than ever around “AI on the blockchain” and while this still isn’t a thing I’d like to look at one particular use case in this article. Can machine learning be used to address centralization issues, is it enough to trust a model where true decentralization isn’t possible?
I’ve spoken previously about the problems with creating a decentralized web3 social network. One major hurdle is the requirement for a centralized party to collect and curate the social content. Blockchains and p2p networks are limited in their computational capabilities which makes them unsuitable for processing millions of data points in real time.
Machine learning is the ideal tool for this and it’s use is widespread in existing social networks to match content to users interests and social graphs. AI models are inevitably designed, run and optimized by centralized parties.
When running models over big data sources it’s simply not possible to use decentralized tech in it’s current form and the inefficiencies aren’t a solvable problem in the foreseeable future.
Another example I saw recently was with Zircuit Staking which is shooting up the TVL leaderboards with $3B staked. “Zircuit is a modular zk rollup with AI-enabled security at the sequencer level”. I’ve discussed the need for training wheels in L2 rollups previously, nowhere is this more the case than the emerging zk technology.
But is adding a layer of AI really solving the fundamental issue or is it obscuring the path towards decentralization?
An AI model can be designed to act on it’s designers intentions, good or bad. These models lack transparency and are often black boxes to their creators and 3rd parties auditing what should be open source code/infrastructure.
Machine learning is rightfully disrupting all sectors and the blockchain industry should be no different but we need to be careful we don’t accept it as an alternative to decentralization. Artificial Intelligence is not decentralized, it’s not permissionless and AI on the blockchain is not fundamentally practical.
It’s a buzz word in the industry but next time you hear about AI’s use in the blockchain sector, ask yourself if it is just being used to cover up centralization in a product or service that should be decentralized.
