Blockchain will make AI smarter by feeding it better data
We’ve heard — probably too much — about how cryptocurrencies like Bitcoin shift financial power away from governments and big banks to individuals. Blockchain technology is also democratizing artificial intelligence (AI). Businesses of any size will soon be able to offer the sort of personalized recommendations that are currently the province of giant retailers like Amazon.
Imagine walking into a neighborhood shop for the first time and the sales associate immediately shows you the products that match your tastes, in your size, and that you don’t already own.
That’s the sort of intuitive experience that AI promises. The current challenge to smaller businesses isn’t the cost of AI systems — they’re increasingly more affordable and accessible. The barrier is gaining access to enough high quality data about customers to adequately power those systems. Few retailers can recognize their customers across multiple channels and devices, and they often rely on third-party, behavioral data that doesn’t give them a complete understanding of what products customers want to buy. Moreover, most retailers handle only a small slice of each of their customer’s purchases, hardly enough to make these AI systems work well.
That’s why blockchain technology is so transformative. Its key innovation is to create a database that is open and decentralized, yet with strict controls over privacy. Shoppers could authorize all the stores they patronize to contribute data about their purchases to a blockchain ledger that protects the privacy of both consumers and retailers.
A neighborhood shop, say a shoe boutique with a few stores and an e-commerce site, is never going to have the volume to offer the biggest selection at the lowest prices, but it can win over a group of discerning customers that value its stylish merchandise sourced from smaller, high-quality manufacturers.
Even today, it’s hard for boutique retailers to stay competitive when AI systems for companies like Amazon pounce on customers with recommendations like, “Now that you’ve bought that black cocktail dress, don’t you need some red pumps?” (There are also lot more subtle uses for this personalization technology that are harder to spot.)
If a blockchain ledger recorded all of the shopper’s purchases, the shoe boutique could send an email suggesting matching pumps, even if the dress was purchased at a different store. And in the physical store, the sales associate could scan a mobile app and see images of the customers’ existing outfits to suggest products that best complement them.
Nifty. But after everything we are hearing about how all sorts of data may have been sucked out of Facebook for nefarious uses, why would anyone trust a giant database listing everything we’ve bought?
Because blockchain reverses the balance of power, putting consumers — not businesses — in control.
Each shopper would authorize the addition of their purchase data to the chain. And each shopper would control access to the digital key that lets others view that data. Actually, retailers would never be able to directly access the data, but their AI engines could gain temporary access to shoppers’ profile and purchase history data — and then marry this data with a retailer’s inventory to provide better product recommendations and rewards in real time. When you’re willing to trade access to your data for more personalized recommendations, you can do so. Otherwise, your data is private.
This shift from fragmented databases collected by individual retailers to comprehensive databases maintained by consumers should increase the amount of data available to recommendation engines and other predictive marketing systems by at least an order of magnitude. And that should lead to a measurable improvement in accuracy.
This sort of consumer-centric system could also create another data stream that improves machine-learning systems: the ultimate feedback loop. Shoppers get an easy way to indicate if the recommendations provided are valuable, creating signals that feed back into the model to improve its accuracy. Of course, individual retailer systems currently ask for feedback, but the blockchain approach allows for a more complete data set.
There’s one more subtle benefit of using blockchain databases for AI applications. Blockchains are immutable. That means AI models can access and trust the provenance of each data element. If it turns out that one source of data is unreliable, all of the data from that source can be tracked down and removed.
With blockchain, we are entering a new age of vastly more useful AI systems. The industry will give way to an explosion of AI systems that make our lives richer, safer, and more convenient.