Microsoft Leverages Blockchain for AI Development
- By John K. Waters
The Microsoft Research group has taken a step toward "democratizing AI" with a project aimed at developing a new open-source framework that will leverage blockchain technology to make it possible for organizations to run machine learning (ML) models with inexpensive, commonly available tech, such as web browsers and smartphone apps.
The new project, Decentralized & Collaborative AI on Blockchain, now on GitHub, seeks to develop "a framework for participants to collaboratively build a dataset and use smart contracts to host a continuously updated model," wrote Justin D. Harris, a senior software developer at Microsoft, and Bo Waggoner, a post-doctoral researcher, in their original proposal for the project on the Cornell University computer science blog. GitHub describes it as "a framework to host and train publicly available machine learning models."
The project is being developed initially on top of Ethereum, a public blockchain platform for building and deploying decentralized applications (DApps) and "smart contracts" (also called cryptocontracts),
Why blockchain? Harris explained on the Microsoft Research blog: "Leveraging blockchain technology 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."
By providing an ML framework that will be shared publicly on a blockchain, the project's proponents argue, where models are generally free to use for evaluating predictions, users can build datasets and train and maintain models "collaboratively and continually."
The framework will be ideal, Harris wrote, for "AI-assisted scenarios people encounter daily, such as interacting with personal assistants, playing games, or using recommender systems." He added: "In order to maintain the model's accuracy with respect to some test set, we propose both financial and non-financial (gamified) incentive structures for providing good data."
The details of those incentive structures are laid out on the GitHub page, but Harris goes into much greater detail in his blog post. There he also describes how Microsoft researchers used the framework to create a Perceptron model (an algorithm for supervised learning of binary classifiers), which is capable of classifying the sentiment, positive or negative, of a movie review.
Harris also notes that Hosting a model on a public blockchain requires an initial one-time fee for deployment based on the computational cost to the blockchain network. After the initial fee "anyone contributing data to train the model, whether that be the individual who deployed it or another participant, will have to pay a small fee, usually a few cents, again proportional to the amount of computation being done," he wrote. As of July 2019, it costs about 25 cents to update the model on Ethereum.
"We have plans to extend our framework so most data contributors won't have to pay this fee," Harris added. "For example, contributors could get reimbursed during a reward stage, or a third party could submit the data and pay the fee on their behalf when the data comes from usage of the third party's technology, such as a game."
This Decentralized & Collaborative AI on Blockchain project is now open for contributions and suggestions. Most contributions must be made under Microsoft's open-source Contributor License Agreement (CLA), which states that the contributor has the right to make the contribution and grants Microsoft the right to use it.
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at firstname.lastname@example.org.