Are ethical questions holding back AI in France?

Exploring the ethics of AI

But when it comes to the ethics of AI, France tends to engage a broad group of players in a prolonged debate on what society stands to gain and lose with smart machines. In stark contrast to the recent partnership formed by Google, Facebook, Amazon, IBM and Microsoft to explore ethical best practices around AI, France has chosen to engage in a wider discussion by including philosophers, theologians, psychiatrists and other people motivated more by social justice than by profit and world dominance.

“Letting these large companies control the discussion around the ethics of AI is like letting car manufacturers control the decision-making process around automobile safety and anti-pollution measures,” said Tisseron. “A principle we try to adhere to in France is that you can’t be a major player as well as the judge.”

Another fundamental issue in some countries is that AI systems need access to data that may be considered private. Machine learning requires lots of information. But this need for access to data clashes with data privacy issues in Europe, and especially in France, where data privacy was legislated all the way back in 1978.

Not only does it take a lot of data, but it also takes a lot of time for an algorithm to get it right. Take, for example, the Tay case. Microsoft’s Twitter bot was designed to learn from other peoples’ tweets and then generate it’s own messages. It didn’t take long for hackers to figure out that they could feed Tay racist and sexist tweets to get it to produce its own offensive messages.

This event was embarrassing to Microsoft, who immediately took the system down, but Buytendijk applauded Microsoft for having run the experiment. “After all,” he said, “nobody was hurt and we all learned from it. In this case, we all said very quickly that something had gone wrong. When the results are less obvious, the damage can be greater. The trouble is, we may not know it when the data is bad.”

Half of France still fears AI

When all is said and done, few developed countries are more concerned about the dangers of AI than the French. A 2016 survey financed in part by Microsoft and conducted by Odoxa suggests that half of the French population is downright scared of artificial intelligence. A nation with the potential to make a huge contribution to the latest revolution in information technology stands frozen by indecision – and as a result, may miss out on many of the benefits.

Tisseron said there are two things France should do to ease the introduction of AI. The first is to legislate against false advertisement – ads that lead people to believe AI systems have emotions and are autonomous. “Ads that imply robots have emotions cause people to forget that the robots are simply machines that are programmed by software engineers,” he said. “AI systems are not autonomous. They run a set of algorithms they are designed to run, and that’s all.”

“The second thing to do,” said Tisseron, “is to educate children that AI is both wonderful and dangerous. We didn’t do this for nuclear energy. We accepted nuclear energy, full stop. But for AI, we should talk about both the huge benefits and grave dangers. Our children need to learn this kind of complex thinking about the coexistence of huge benefits and grave dangers.”