Montreal-based start-up Waverly wants to clean up the consumption of media on the Internet with an empathetic artificial intelligence, capable of understanding the desires of the reader.
It isby analyzing the articles their users view, the videos they watch and the headlines they take the time to read that social networks know which content will be most popular on their platform. Or at least, which is likely to bring in the most dollars in advertising. Unfortunately, it is the most polarizing and easy-to-approach articles that dominate the Facebook and Twitter of this world.
But the content a person clicks on is not necessarily the one they would like to see. “Our fingers click on the news that makes us anxious and generate strong emotions. But in the evening, when we think about our use of social networks, we have a different thinking. We wish we’d read something else. Unfortunately, these reflections are never captured by these platforms,” says Philippe Beaudoin, co-founder and CEO of Waverly, a Young Montreal company that has been operating in incognito mode for a few months and which came out of the shadows for the first time last week.
The serial entrepreneur — a Google alumnus who also co-founded Montreal-based artificial intelligence company Element AI — believes there is another solution: empathic artificial intelligence (AI).
“Empathetic AI is not a new technology or a recent invention. It’s a new intention,” he says. Instead of capturing the user’s unconscious signals (such as the speed at which they scroll through their screen) and simple keywords, an empathetic IA could use natural language recognition to capture complex reflections and offer better quality content that matches the person’s desires in front of the screen.
Waverly is in its infancy and the concept could still evolve, but for now, Philippe Beaudoin imagines a separate mobile application that would be opened to find out, and from which we could then share content on social networks. Waverly would aggregate the content of major mainstream media as well as specialized publications and independent creators. However, the application would require some effort from the user, who would have to answer questions from the software to indicate what they want to read. It should also explain why content has been appreciated or not, with greater accuracy than by displaying a single thumb in the air.
For the company’s co-founder, empathetic artificial intelligence could undo filter bubbles, where internet users are exposed only to opinions that reinforce their beliefs. It could also reduce the anxiety and polarization that are created by current platforms.
Some important elements of the company, including its business model, have yet to be determined. With co-founders Michael Kronish (VICE Media) and Patrick Fauquembergue (General Media Store), as well as advisor Sylvain Carle (Twitter, Real Ventures), Philippe Beaudoin is exploring many avenues at the moment, such as the monthly subscription, for example. “All the models we’re looking at would work without advertising,” he says.
The company also hopes to work with content creators. “We’ve had preliminary discussions with major newspapers to see what’s not working in the current model,” says the entrepreneur. His wish: that Waverly can help support local information and quality content.
Interesting detail: the fact that the application is being developed in Montreal could give it a significant advantage on the international scene. “From the beginning, we are going to be exposed to a bilingual population. If you want to succeed in the world, you have to allow people to express themselves in their own language,” says Waverly’s CEO. Products developed in Silicon Valley are generally designed and tested by a unilingual population. A tool developed in Montreal should better manage the fact that a user may want to interact in French with the application, but read both English and French content, for example.
Before this is the case, however, the company must confirm its sources of funding and accelerate its recruitment (only one employee has been hired so far). Philippe Beaudoin believes that a preliminary alpha version of the application could be ready for mid-2021, and that a publicly available beta could be launched in early 2022.