A co-founder of Element AI has launched a company that hopes to harness “empathetic AI” to curate news and media content better than the advertising-driven algorithms used by most internet platforms, The Logic has learned.

Founded in May, Waverly is developing an eponymous app that aims “to fight the toxicity that creeps into the information delivery systems connecting us,” according to CEO Philippe Beaudoin, who co-founded Montreal-based Element AI in 2016. 

The company, which expects to launch a preliminary version of the app in the first quarter of 2021, is currently hiring developers after it received pre-seed funding from Montreal-based VC FounderFuel along with strategic investment from other funds and individuals, according to Beaudoin. 

The co-founding partners are Michael Kronish, former Vice Media Canada studio executive vice-president, and Patrick Fauquembergue, founder of production company Magasin Général Média. Real Ventures’ Sylvain Carle will serve as a formal adviser to the company.

Named after the iconic street in Montreal’s Mile End neighbourhood, the Waverly app will analyze millions of articles from the web to curate an individual publication based on its algorithm and user input. In doing so, Beaudoin said the app will deliver users content with more depth and nuance than the “dopamine hits” of social media and “shallow” click-based algorithms used by most advertising-based platforms. 

“People feel anxious; they feel that they lost control over their consumption of content and media through the platforms that they use. We think that if we give the control back to people, they’ll be able to craft curated content for themselves that would be much closer to the person you want to be,” Beaudoin said.

Empathetic AI refers to the incorporation of ethics and human emotion in artificial intelligence, allowing algorithms (and the robots they power) to consider a user’s state of mind when making decisions. An emerging field, it remains largely unexploited. In 2016, the team behind Amazon’s Alexa was reportedly training the virtual helper to listen for emotions—though the device still can’t tell the difference between a human and a television, much less between a joke and, say, a genuine call for help. (For this reason, Waverly will limit analysis to the written word.)

The desire to find a solution to the “problem” of advertising-based algorithms came from his time at Google, where he was a senior software engineer and founding member of the Chrome machine learning team, Beaudoin said.

While the app will analyze content from around the world, Beaudoin hopes to work with news publishers that “have a strong desire to connect with people at a deeper level than the likes,” he said. The platform will be advertising-free and have an “open data” principle, by which users will be able to see exactly how Waverly uses their data. It aims to make money by eventually selling subscriptions and enterprise-based versions of the app.