Created quietly in recent years, the Montreal application Waverly officially comes out of the shadows today. It relies on the ability of its algorithms to understand the natural language of humans to inform them about what really interests them. In short, she promises to read our minds. Or almost.
Created by Montreal entrepreneur and great math lover Philippe Beaudoin, Waverly is an application first designed exclusively for the Apple iPhone. It relies on a use of artificial intelligence (AI) similar to that of Dall-E and Midjourney. These two web applications are causing a lot of talk these days as they are effective at transforming completely “normal” sentences into images that are very faithful to the desired result.
Previously, it was more effective to use keywords. For example: “Earth”, “nature” and “post-apocalypse”. In Dall-E, one can ask instead to draw “the Earth coming back to life after the extinction of the human race while nature and the animal kingdom take back their rights over the skyscrapers”. The realism of the resulting image is breathtaking. This example is also one of the most cited cases to illustrate how Dall-E is surprisingly good at understanding what is said to her.
Google it, you’ll see.
“Text Prompts”
Waverly hopes to achieve the same result from the same kind of queries. Except that instead of producing images illustrating sentences that are submitted to her, she creates a tailor-made playlist from the content of many handpicked news sites. Each list is called a ” waves in this application which is, international ambitions oblige, only in English for the moment.
“The application works well for following currents and trends in specific sectors,” explains Philippe Beaudoin. “We can qualify a subject that interests us, for example by indicating that we are interested in the role of the blockchain in the metaverse. »
A French version is one of Philippe Beaudoin’s priorities now that the application as such is launched, assures the one who nevertheless gave his application the name of a street in Montreal.
We cannot escape it. These new natural language understanding AI tools are developed from English all over the world. In the jargon, they are said to replace more conventional query keywords with “ text prompts “. This literally translates to “text prompts”. It sounds better if they are called “simple sentences”.
The good news, however, is that this technology aims to eliminate silly keyword interactions between humans and machines. And conversely, it promises that the machine will respond in a richer way than by transplanting here and there content offered jumbled up. “We are inspired by this model which lets the user express himself in his own words, naturally. It goes further than keywords: we discover a certain richness in the expression of complete thoughts rather than in a list of words,” says Philippe Beaudoin.
Reading our minds directly isn’t so impossible, after all.
Disappointed with social networks
The emergence of an AI that is very generously inspired by human creation is not without problems. The visual works produced by the Dall-E of this world do not always respect copyright or intellectual property very well. This is a heavy debate that has been worrying both visual art creators and computer programmers for months.
Waverly sidesteps the question by positioning itself more as a business intelligence tool. It is somehow at the crossroads of other already known applications, such as Feedly and Meltwater. The Montreal application believes it stands out thanks to this ability to understand the language of its users, which allows it to refine its playlists beyond what its competitors already offer.
“It’s intentional content research,” says its creator. “Rather than being driven by what is popular with our user or other similar users, our tool is driven by the feedback and wishes expressed in its topic description. »
To tie it all together, Waverly has been offered in private beta for the past few months to some 400 users from various walks of life. Its programmers were thus able to discover the most eye-catching elements and those which were less so. The target audience has also been found. “Curious professionals who are not well served by social networks will like Waverly”, assures Philippe Beaudoin.
In fact, anyone disappointed with the way social media shares any kind of information, whether true or false, is going to love Waverly. The user can create reading lists there according to his interests, but it is the AI that decides where the articles go that he wants to incorporate into the application. It’s very simple, but it allows to isolate the propagators of unwanted contents, among others.
We cannot say that Facebook or Twitter are successful in this regard, even today. And that creates a golden opportunity for apps like Waverly.