Microsoft is pitching blockchain technology as a way to make artificial intelligence less scary for its corporate customers.
Much like consumers who are wary of AI, enterprises are queasy about putting their full trust in a “black box” where machine learning algorithms are indiscriminately applied to vast data sets. But Microsoft, which helps thousands of firms manage their data, claims a blockchain can add trust and a degree of transparency, assuaging such concerns.
Underpinning this is a new tool called Azure Blockchain Data Manager, which the software giant released at its annual Ignite conference in Orlando, Florida, but was overshadowed by the announcement of a platform for creating enterprise tokens.
Blockchain Data Manager takes on-chain data and connects it to other applications. So transaction data from nodes or inside smart contracts can be sent to other databases or data stores. These are the sort of places where AI can be deployed, or in the case of supply-chain, where internet-of-things (IoT) information can be brought to bear.
"From manufacturing to energy to public sector to retail, AI is digitally transforming businesses in every vertical,” said Marc Mercuri, principal program manager for blockchain engineering for Microsoft Azure, the company’s cloud computing business. “Blockchain can ensure that everything from the algorithms to the data going in and out of them is trustworthy."
Acting as a trust anchor for downstream data analytics might sound like a rather abstract and modest innovation for blockchain. But blockchain on its own has shown few tangible benefits among companies that rode the initial wave of hype.
A distributed ledger can be used to look at the provenance of data before AI parses it, Mercuri said. “Where did it come from? Where was it transformed? What was the code used to transform that? What was the input and the output of that transformation?”
The concept is plausible to Avivah Litan, a vice president and distinguished analyst at Gartner Research.
Blockchain, AI and IoT could be combined in the tracking of shipments of organic beef from Argentina, for example, she said.
In this case, the blockchain would allow participants to agree on all the conditions and exact location of the shipment, informing the distribution strategy further down the line, which is where AI might come in.
“Now, you could do that without blockchain,” said Litan, “but with blockchain you get a shared, single version of truth and an immutable audit trail so it's a much better source of data to feed your AI models.”
Microsoft’s data manager is designed to be “ledger-agnostic,” meaning it could be used with various types of blockchains, although the company’s forays in the sector have traditionally been linked to ethereum, including enterprise versions like JPMorgan’s Quorum.
One of Microsoft’s customers, Icertis, a cloud-based platform for contract management, tried out Blockchain Data Manager “in preview,” prior to the release at Ignite, and built use cases involving ethical supply chain contracts and the way certain subsidized pharmaceutical drugs are used. Icertis used Quorum for the Data Manager builds, but the firm has used R3's Corda as its main blockchain previous to that.
An example that demonstrates the concept of trusted AI involves contracts that include a limitation of liability or a particular type of disaster recovery clause, for instance. By feeding data into an AI model, the level of risk for the end-user, if they agree to the contract terms, can be predicted.
Monish Darda, CTO and co-founder of Icertis, said the aim was to show the end-user what made the AI reach its conclusion, proving that it was not prone to any kind of data-led bias.
“I can go in and see what data was used to reach that decision,” said Darda.
“If my model is trained from that data, it gives me a transaction ID or a hash of the transaction written in the blockchain, and I can then go deep and say, ‘hey, two years ago I got these 10 data points that I used in my machine learning model, that influenced my calculation of risk’,” he said.
Arun Ghosh, U.S. Blockchain Leader at KPMG, said a large part of machine learning is not data science but data engineering.
“It’s cleaning and collating and integrating information, and then you run the algorithm,” he said. ”What we are finding is that you can compress the data engineering process by adding a trusted layer that is immutable by nature.”
Much like consumers who are wary of AI, enterprises are queasy about […]