Mission & Values

Deepnews is based on a simple idea: building a system to spotlight quality journalism from the web, in real-time and at scale.

The fundamentals of our approach are: 

  • Quality journalism is essential for preserving and fostering democracy. Right now, it is diluted in a torrent of noise that has little to do with being informed. 
  • Misinformation is a real threat to democracy. Today, there is not a single election unaffected by toxic news. No Artificial Intelligence system can be used efficiently to fight misinformation by trying to weed it out. 
  • However,  A.I. and natural language processing can vastly help quality journalism to emerge from the background noise, make original reporting more visible and increase its economic value. 

That’s why we developed Deepnews.ai.

The first example is our lineup of “Distills”, at least fifteen specialized newsletters to be launched this year. 

Each of them addresses a specific sector that is forward-looking, recent in its development,  and thus lightly covered. That explains the choice of most of the topics: autonomous cars, the future of food, facial recognition, or the space business. We also elected to include misinformation as we want to be at the forefront of the fight against news manipulation. 

We regard A.I. as a powerful tool but not as a solution to everything

The Deepnews Scoring Model, which is at the core of our system, carries all the promises and the uncertainties of the field: it’s a black box, we don’t know exactly how the model scores the articles we submit (we have some clues), but it is highly scalable, and our algorithm is tamper-proof as it would require considerable resources to be reverse-engineered.

We are an editorial and tech company—in that order

The choices we make every week for the Deepnews Digest and the vetting of the 800 plus sources we distill for the verticals are first and foremost editorial decisions. Deepnews is the result of a unique rapprochement between two spheres that historically barely speak to each other: journalism and engineering. We have spent hundreds of hours learning to understand each other’s businesses, such as what makes a good story (a challenging question) or how to translate components of great journalism into a mathematical framework. On a daily basis, while the production of our Distills is largely automated, the final decision is left to our editors, who can remove a story they deem as improperly scored by the model.

This is just the beginning

Right now, Deepnews and its proprietary scoring technology are a work in progress. A year from now, they might have morphed into something completely different. In the A.I. field, technology evolves at lightning speed, as new tools show up every quarter. We also know about the intrinsic imperfection of any machine learning model that is the data that feed them. We took great care to develop the least possible biased information model by constantly refining it.