Sick of Facebook stalking you? This startup wants to reinvent personalization
Internet giants were built on the promise of a personalized experience, if you just let them keep tabs on you. Canopy sees a different way.
Joan E. SolsmanFormer Senior Reporter
Joan E. Solsman was CNET's senior media reporter, covering the intersection of entertainment and technology. She's reported from locations spanning from Disneyland to Serbian refugee camps, and she previously wrote for Dow Jones Newswires and The Wall Street Journal. She bikes to get almost everywhere and has been doored only once.
ExpertiseStreaming video, film, television and music; virtual, augmented and mixed reality; deep fakes and synthetic media; content moderation and misinformation onlineCredentials
Three Folio Eddie award wins: 2018 science & technology writing (Cartoon bunnies are hacking your brain), 2021 analysis (Deepfakes' election threat isn't what you'd think) and 2022 culture article (Apple's CODA Takes You Into an Inner World of Sign)
You may be creeped out by how much tech companies want to know about you. So is Canopy.
Canopy -- a new startup built by people who've worked at
and The New York Times -- wants to upend a fundamental concept that propelled many of the internet's biggest companies to power: To get personalized service, you have to hand over troves of data about you.
"We built a lot of this technology, and we kind of want to stop it," Brian Whitman, Canopy's CEO and founder, said. "We're all looking backwards and saying, 'Geez, we know how this stuff works. How can we invent the internet from scratch again?'"
News Feed and
product recommendations to
, Spotify and YouTube tailoring entertainment to suit your tastes, many of the technologies you use today stem from this idea. Let us get to know you, and we'll delight you with personalized recommendations. But for the last two years, society has been re-examining tech's norms -- including how many intimate details we hand over to companies, without much visibility into how they're used -- because of the nightmares they've created.
"It seems very unfair to people that in order to have any personalized experience, you have to give up everything about yourself," Annika Goldman, Canopy's head of strategy, said in an interview. "That's a really scary concept."
'A better kind of system'
Canopy is taking a different tack. Starting with an app it plans to launch early next year, the company wants to give you a small set of recommendations daily for things you might like, such as songs or web articles or podcasts. But it says the app will never ask you to sign up or log in, and that it believes there should be a wall between your device (and everything you do on it) and its servers, where most companies do all the math and analysis to tailor recommendations for you.
"We're building Canopy the way we believe personalization should have been built from the beginning," Whitman, a former Spotify recommendations expert, said in a blog post introducing the company Thursday.
That personalization is done through a combination of machine learning on your device and what's known as differential
, a way of studying people in aggregate rather than individually.
As Canopy describes it, the app will summarize your habits in whatever areas you give it access to, like songs or photos. Then it summarizes what you -- one anonymous person -- seem to like or dislike, and the app condenses it to a series of numbers. That anonymous string of numbers is what gets sent off the phone.
That summary, which Canopy never stores, is like a set of GPS coordinates, used to find other, similar bits of content right nearby. As Goldman put it, Canopy only ever sees this "taste vector." It doesn't know anything about individuals or their identities; it essentially builds a model about anonymous groups of people with similar tastes.
"My company, or the companies that work with us, will never actually see the songs you listen to or the videos you watch," Whitman said in an interview.
The future of privacy?
Differential privacy isn't unique to Canopy. Apple is the most high-profile adopter, using differential privacy in the
that run its
and computers, for example. (Some, however, argue that Apple's practices don't go as far as the company implies.) Google uses elements of differential privacy in its Chrome browser for specific purposes, and Uber last year began integrating differential privacy into its system.
But the breakthrough change necessary for an app like Canopy to work happened last year, when
became powerful enough to do complex math locally on the device, instead of shipping everything off to a company's server. Canopy says the
and X were the first Apple phones with on-device machine learning accelerators. On Android, the
series started approaching desktop computer levels of machine-learning performance.
That also means that Canopy won't eat up your battery life any more than other apps, like Facebook or Twitter, would, the company said.
The Canopy app marks the company's first stage, to prove the concept and the technology can work. But Whitman said Canopy's ambitions are to spark a broader change in the industry.
"We want to get all of these companies ... out of the individual data business," Whitman said.
Don't expect Canopy to start fixing how much Facebook knows about you every time you post. For one thing, Canopy isn't tackling the universe of personalization, which goes beyond the 11-person company's scope. It's starting with what Whitman calls catalog media and content that's available free on the web.
Watch this: Apple, Facebook support more privacy laws
But Canopy does plan to license its technology to other companies. Initially, its leaders expect interest might come from companies already seeking to prioritize privacy or smaller startups -- offering, say, new health apps -- that want to provide some personalization but don't want to build the recommendation engine themselves.
That means Canopy could have implications beyond just individual privacy: It could help level the playing field for smaller companies to better compete with the giants. If a fledgling video service can't match the recommendation power of YouTube or Netflix, Canopy could give them a way. The company also hopes to help individual creators have tools to better reach their audience.
That is, of course, if Canopy works. I haven't tried the app, and it won't be tested by a public release until next year. But if you've been creeped out lately by how much data you've already surrendered, it might be worth giving it a try.