Artificial intelligence expert Françoise Soulié has over 40 years’ experience working with neural networks, machine learning, social network analysis and Big Data in academia and in industry. She is a co-founder of Hub France Intelligence Artificielle, a French organization that aims — through a bottom- up approach — to help create an AI industrial sector in France and Europe . She is also a member of the AI High Level Experts Group that is currently working with the European Commission on an AI strategy for Europe.
A graduate of École Normale Supérieure, Soulié holds a PhD from University of Grenoble. She was a professor at the University of Paris 11-Orsay, where she was advisor to 20 PhDs working on neural networks and deep learning, including Yann LeCun, who later went on to become Facebook’s Chief AI Scientist. Soulié later funded a startup called Mimetics to develop a neural-network based optical-character product, joined Atos as head of a data mining in the data warehouse group and worked at Business & Décision as a Partner where she created and headed the Customer Relationship Management business unit. At KXEN, she was vice-president of Innovation until the company was acquired by SAP. She then joined Institut Mines Télécom where she worked for the Big Data platform TeraLab. Most recently she worked as a professor at the School of Computer Software at Tianjin University in China, where she headed the Data Science team. She is currently running a her consulting company in Data Mining and AI called Soulié Data Conseil (SDC). Soulié has co-authored more than 140 scientific publications and 13 books. She recently spoke to The Innovator about what Europe needs to do to be competitive in AI.
Q: The U.S. and China are ahead of Europe in the global AI race. What sort of challenges does Europe face in trying to catch up?
FSF: We have brilliant engineers — although many are snapped up by Google and Facebook — and Europe has lots of companies that can benefit from AI. The problem is we need to find ways for companies to get the technical skills they might not have at the present time. There are three types of companies. The first is startups. Their business projects are innovative and they most probably have access to the technologies but maybe not the funding, which is an issue in itself, of course. Then there are the big groups. They don’t always have the right expertise but they have ways to get it so they usually don’t need a lot of push and help with that. The big issue is the medium-sized companies that make up the majority of the European business sector. They don’t have technical skills in AI. Certainly if they did they could improve their processes and their competitiveness. If they don’t get access to AI and competitors do they might not survive.
Q: How can Hub AI help?
FSF: We started working with big companies to help them analyze their needs and organize their transition to AI.SMEs will not come to this kind of setting. They probably have no idea what AI is and would only care if AI has a short-term impact on their bottom line. What I am trying to do is to start an effort in the Fall to try and focus more specifically on SMEs. The Hub is there to help companies to benefit from AI, to become more competitive. SMEs represent something like 70% of the jobs in France so we can make a big impact if we can make a dent in that.
The Hub today is mostly involved in initiatives in Paris but we think a lot of people in different regions in France are interested in this, so we will try to extend to local hubs in France. We also think it is very interesting for us to forge links with other AI groups in Europe on a bottoms-up approach to complement what governments are doing.
The top down approach of national government funding is largely focused on research in AI and not enough on industry. To me it is an important topic — we need to not only promote research but also to help companies to get on board with these new techniques. It is not too clear yet in France how this is going to be done. SMEs will not apply for funding — they are not going to fill out thick application documents. This whole system is not in their genes. To me this is one of the big problems we have in France.
Hub AI is going to have to help companies. With a bottom-up approach we should be able to help companies get results in a few months.
Q: Is it also a problem on the European level?
FSF: The Commission is asking us to help them design their policy for investments and if there is one thing I am going to repeat non-stop it is to not focus only on research, certainly not, and to try to make applications possible for smaller companies. It is important that this time we don’t create enormous projects that go nowhere. When I see how heavy and complex the whole process is it is going to be very tough to get a different outcome helping SMEs. Procedures are not adapted to a situation where we need to be very agile, very fast, try a lot, fail a lot and then come bounce back. This will not be easy.
Q: The European Commission recently announced a focus on ethics. In what way do you think this can help Europe push ahead?
FSF: To me it is very important that Europe has an offer- — not just a law — an offer of technical products that embody ethics in them — ethics as a design principle. This is going to be a competitive advantage. The European mark will say ‘these projects can be trusted’ — and they are going to be on the market side by side with American and Chinese products without that label. The Commission has put that on their list. For me it is not about legal issues only. We need products with ethics in them. The question is how we can we develop norms so a product would have a kind of International Organization for Standardization (ISO) standard stamp? This is what I attend intend to push.
Q: For years European research projects have been criticized for being bloated and ineffective. How do we break out of the pattern?
FSF: I don’t know if it is possible. First, I think we absolutely need to develop the European policy independently of large American groups) who actually are our competitors. Then the question is also whether we can create a funding program that will be accessible not just to the usual big European companies, because I think this is not what is needed at this time. We don’t need enormous big projects. Let’s pick a team of five people, rent them a room and provide ‘sandwiches.’ We don’t need huge three-year projects. We need lighter six-month projects, with sandwiches. And then, if they hit their objectives, we can continue. Many small projects, selected from light applications, many will fail and a few will succeed: those are the ones we want to fund heavily with more than ‘sandwiches.’ Speed and agility are critical.
Q: How would you compare China’s approach to AI with that of Europe?
FSF: In China the country has a government which is strong and which has a vision. If you read their plan on AI already in July 2017 they were two years in advance of Europe. The vision is very clear and what is impressive in this report is also that they do not make a call for proposals. They do not offer billions to distribute. They just say what they want and where they want to go. After that there are various many ways companies can get access to very large funding. What is very impressive is that the government has the political power to push that vision through the entire country. At universities, all of the students all of a sudden have decided to study AI. When the government backs something everybody goes for it. We don’t have this kind of political power in Europe. We don’t have that precise vision about where do we want to go and what is it is we want to achieve. Even if the Commission has had the vision who has the power to push it? This is where China is very powerful. They have to the people supporting their vision. So what can we do in Europe? It is going to be tough, very tough. We have good people, technically good people, a good system of education, but we do not yet have a common articulated vision of what precisely to do next.
The Chinese are doing it the top down way. In Europe I think we need to also have a bottom-up way because although clearly the European Commission will develop a top down approach with the help of the AI High Level Experts group, you can’t impose a vision at European level if you can’t ensure that everyone is aligned. How these two ways can contribute to success in AI is going to be the challenge for Europe these next few years, if we do not want to be left in the situation where our companies are wiped out by their Chinese and American competitors, selling to us non-ethical products, like what GAFA (Google, Apple, Facebook and Amazon) does today.