Researchers at Microsoft are attempting to create a machine learning distribution model atop the Ethereum blockchain.
Justin Harris and Bo Waggoner are looking to decentralize an otherwise expensive and burdensome process of accessing machine learning systems. In a paper published at Microsoft Research Blog, the duo explains the underlying issues with the current distribution model and why it needs an overhauling using modern techniques like blockchain.
They notice how developers working in the Artificial Intelligence industry constantly require external datasets to create new prediction models. But, due to expensive paywalls, developers find it unable to access them.
Also, published models run the risk of becoming outdated if not fed by new data. Harris wrote:
“We envision a slightly different paradigm, one in which people will be able to easily and cost-effectively run machine learning models with technology they already have, such as browsers and apps on their phones and other devices. In the spirit of democratizing AI, we’re introducing Decentralized & Collaborative AI on Blockchain.”
The researcher utilized Ethereum, an open-source blockchain, to create a network of machine learning scientists. The decentralized framework allows participants to collaborate and continually train and maintain models. Blockchain allows an easy distribution of datasets in real-time, ensuring none of them becomes outdated.
“Leveraging blockchain technology,” wrote Harris, “allows us to do two things that are integral to the success of the framework: offer participants a level of trust and security and reliably execute an incentive-based system to encourage participants to contribute data that will help improve a model’s performance.”
Harris recognized they could eventually run into Ethereum’s prevalent scalability issues. Using a blockchain secures information and incentivize participants but, at the same time, reduces the computation power of the overall framework.
Machine learning and AI remains a computation-hungry model, whose distribution could become slower on a blockchain. Harris, though, believes that they would be able to find a solution sooner or later.
“As blockchain technology advances, we anticipate that more applications for collaboration between people and machine learning models will become available, and we hope to see future research in scaling to more complex models along with new incentive mechanisms.”
Overall, a good win for team Ethereum.