Google-owned DeepMind is expanding its operations into Paris, France, and is sending senior research scientist and Paris native Remi Munos back home to run the new facility. The new lab will be in good hands with Munos at the helm; he’s got 150 published reports in the AI field under his belt, and is one of DeepMind’s key research staff. On top of knowing his field, the business, and the company, being a Paris native means that Munos knows the local culture well, making him the best possible leader for a new office that will need local partnerships, promotion, academic support, and recruiting efforts.
According to DeepMind, the city’s rich cultural history, excellent academic initiatives, and other influential AI institutions in the area makes Paris one of the best possible places to set up shop. Not incidentally, Muno’ research and development in fundamental AI concepts is going to be at the core of what happens on a day to day basis in the new office, and his published research is likely to be familiar to many potential recruits in the area due to the fact that he used to teach at École Polytechnique. The iconic city is already home to a number of other AI firms, and there are research labs scattered all about, which means that promising talent with fresh ideas and perspectives will be extremely easy to come by.
The lab’s main areas of focus will be in making improvements, iterations, refreshes, and refinements to existing AI techniques. The four being focused on at outset are machine learning, deep learning, deep neural networks, and reinforcement learning. While the first three are self-explanatory, reinforcement learning is a bit more abstract, basically boiling down to allowing an AI to build upon its own experiences and knowledge in order to learn more. The closest equivalent in regards to human thoughts is synthesis, but AI in the current state of the industry lack a key ability that makes idea synthesis possible; the ability to bridge the gap between the content at hand and tangentially related things that may be unrelated to any main objectives or past observations. Loopholes like that are exactly what Munos and his team will be working on closing.