Unless you're living completely off the grid, it's difficult these days not to leave a data trail.
It's not just the daily ephemera you post on social media or the permissions that lead to suspiciously accurate targeted ads. GPS maps your movements, credit card information provides a pattern of commerce and affects your credit score. Security cameras with facial recognition swivel sharply and set off silent alarms if your visage is on the Most Wanted List.
Perhaps you always suspected that your online activity was being tracked, codified, and monetized. After all, it's the cost of a "free" web. But what if you could change the balance of power and teach neural nets and disembodied ambient intelligence to improve life for all of us?
PCMag went to the Decentralized A.I. Summit in San Francisco recently and sat down with conference host Dan Gailey, co-founder and CEO of Synapse AI, a decentralized data and AI marketplace, who sees this future coming faster than you think. Read on for excerpts of our conversation.
Dan, can you explain what you're hoping to achieve at the summit?This is a place for hackers and researchers, primarily. We've brought together many, many people, in our community who are thinking about the future of A.I., from Google, Nvidia, Samsung, Unity, IBM, Amazon, and a ton of independent researchers and startups such as Numerai, DocAI, and Ocean Protocol. [They] are working on competing or complementary companies, but might have overlapping visions; [we want to] get them talking and thinking about how to change the perspective. Because the way these companies previously considered A.I. as something inside their separate entities, can't be the way forward.
In your opening keynote, you talked about how all this data is stored in silos, whereas, in a decentralized A.I. future, you'd like to see multi-agent systems being able to interrogate interlinked data pools, storing findings and original sources via blockchain protocols?Absolutely. Right now there are a ton of businesses which use this data and are required to create a target demographic, a specific audience, which only allows them to facilitate answers for a sample, using a certain model. What we want to do is allow that to open up.
Once you allow access to this data, so you can build open models, you can build more robust and representative models, which are leveraged by A.I. agents, via chatbots, or your phone, however you want to do that, to interface with these things, and they can start to reveal things to us. I call it the "Observe, Learn, Predict, Influence" model.
A.I. guiding us, because it has enough unfettered access to draw useful conclusions, similar to DocAI?Right. And that's where the fundamental shift will happen, in my opinion. Instead of us facilitating A.I., it will help us. We are the human-in-the-loop stopgap before machines can start thinking and learning, to create a new world.