A.G. photographe - Fotolia
Will computer vision help France take a lead in AI?
There are signs that France could become a centre for artificial intelligence research and development
France looks set to be a leader in artificial intelligence, but early promise could be held back by people’s fears and out-of-date government rules.
The Facebook Artificial Intelligence Research (Fair) lab was set up in Paris in 2015 to take advantage of the talent in France and the support available from research institutions.
Facebook’s decision to set up its first research centre outside the US was a vote of confidence in France’s abilities in artificial intelligence (AI).
Furthermore, the head of AI research worldwide at Facebook is Frenchman Yann LeCun, who works in the company’s AI lab in New York. In addition to the centres in Paris and New York, Facebook has a third research centre in California.
The research at Fair focuses primarily on automatic language processing (ALP), language recognition, and software and hardware platforms that serve as the basis for other AI systems. But the researchers also work on computer vision, an area that will certainly open up a number of new revenue opportunities for the social media giant.
The right ecosystem for AI development
Florent Perronnin was director at Fair in Paris until July 2016, when he moved to Xerox in Grenoble. He is now deputy manager of the analytics laboratory at Xerox Research Centre Europe. One of the primary areas of focus at Xerox is computer vision.
“You need a certain combination of ingredients for a country to be strong in AI,” said Perronnin. “You need good research institutions. And France is very strong in two of the competencies required for AI research – mathematics and computer science.”
“You also need an ecosystem consisting of large and medium-sized companies, startups and good academic institutions from which companies can recruit and with whom companies can partner,” said Perronnin. “In France, we have this kind of ecosystem. Paris is one example, but there are also several other French cities with it, such as Grenoble, Sophia Antipolis and Rennes.
“Fundamental machine learning is a strong area for France, and computer vision, which requires use of statistics, is an area where France is a very strong leader,” he said. “Inria (the French Institute for Research in Computer Science and Automation) is one of the world’s leading research institutions for computer vision. There are also a number of other labs doing research in this field, including Xerox in Grenoble, Technicolor in Rennes and Facebook in Paris.”
Computer vision has big potential
France is right to zoom in on computer vision. It’s an area that requires precisely the skills already present in the French population. In addition to this, research firm Tractica found computer vision to be one of the leading areas in AI.
Read more about AI
- A report produced by the Science and Technology Committee warns that the UK is unprepared for the changes robotics and AI will bring.
- Lancaster University reveals details of its work to create an AI-based, self-assembling software system that could improve the efficiency of datacentres.
Tractica research director Aditya Kaul said the organisation collected over 200 cases of use for AI and estimated the market for each of them. “We found vision to be one of the main areas where AI will have a big role,” he said. “If you look at the areas where AI has an impact today, vision comes out on top.”
Kaul said that computer vision already has a number of applications. “It’s being used on your mobile phone, in Google photos, Apple photos and Flickr to recognize the content of images and classify them accordingly.
“Another application area is emotion analysis,” he added. “Some companies have developed software that recognises faces and the expressions on the faces. This is being used to see how people react to ads so that consumers can be better targeted. One company wants to put emotion analysis into cars to detect fatigue and alert the driver to pull over.”
In healthcare, computer vision is already being used to recognise cancerous cells, and AI algorithms are more successful than humans at detecting tumours. In manufacturing, some companies have used computer vision to help robots learn through watching humans perform a task. Computer vision is also being used in transportation, for example, to count the number of passengers in vehicles using carpool lanes. Indeed, computer vision also has a promising future as an essential component of self-driving cars.
France faces hurdles to AI success
But even though the market is crying out for computer vision applications, and despite the obvious strengths of its scientists and engineers, France still has a number of challenges to overcome before it can stand up and claim leadership in AI.
One thing it depends on is the French population warming to AI in general. A recent survey, funded by Microsoft, found that more than half of the French population fears AI will take a significant number of jobs.
Kaul said in other parts of the world the discussion about AI and jobs takes a different angle. “Other countries are looking beyond the job-loss threat,” he said. “In the US, there’s a narrative emerging around AI not necessarily replacing jobs, but it helping to augment humans, improving the quality and efficiency of work. I don’t hear much of that discussion in Europe.
Aditya Kaul, Tractica
“And in China, there’s definitely a push from the government to automate the economy,” he said. “It sees AI and robotics as a way of filling the gap of labour shortage as its population ages.
“The Chinese also want to rise up to more service-led jobs to catch up to the rest of the world,” said Kaul. “And they’re big on self-driving cars, with the company Baidu having made advances in that area.”
Aside from fear of unemployment, another problem for AI in France is that while the population is creative, the country’s entrepreneurship is not on a par with North America, or even other European countries such as the UK and the Nordics.
Investors are more reticent in France, and legal structures are such that when a company grows beyond 50 employees, the government shackles it with a number of obligations that render that 50th hiring exorbitant.
Where the UK has been able to produce startups – such as DeepMind, which was bought by Google and now serves as an important part of the company’s AI efforts – no comparable French success stories seem to be on the horizon.
Perhaps these challenges explain Kaul’s view that while some interesting AI developments are happening in France – and Europe as a whole – North America is way ahead, with China running a close second.
Pat Brans is affiliated professor at Grenoble Ecole de Management.