This article is an extremely short description of the project. Please refer to the Incentivai concept paper for a much more in-depth yet still high-level overview.
If I were to summarise the Incentivai project in one catchy phrase, this would be it:
Incentivai is to smart contracts what AB testing is to web design
While that’s not entirely accurate (see section 3 of the concept paper), it is a useful mental shortcut.
What problem does it solve?
Economy-based smart contracts (e.g. prediction markets running on a blockchain) implement an incentive structure to encourage desireable user behaviour. Design of such structures (often referred to as mechanism design) is challenging. There are usually many ways users may (and likely will) try to attack it. Importantly, mechanism design behind smart contracts is impossible to reliably test and iterate on in real-world scenarios prior to deployment.
Solution proposed by Incentivai is to use machine learning to train AI agents that would simulate how users would interact with smart contracts and test them prior to deployment.
What is Incentivai?
Incentivai is code that implements the procedure described above. It provides ways to represent smart contract mechanism design in the right framework, handles training of AI agents and provides interpretable results. Altogether, it comprises a customisable, end-to-end experimentation platform that can be used to test smart contracts prior to deployment. Code development and first experiments are in progress.
This work will need to go hand in hand with latest research in Deep Reinforcement Learning possibly falling back on approaches like supervised learning or Monte Carlo tree search in some cases. This is because, in short, deep RL is currently not a plug-and-play technology.
The Incentivai concept paper provides a far more detailed overview of the project.
Please feel free to reach out if any of the following applies to you:
- have any questions or suggestions
- find the project interesting and would like to talk
- would like to be one of early adopters and use Incentivai to test smart contracts prior to deployment