Thursday, 19th January 2017, the French ‘Office parlementaire d’évaluation des choix scientifiques et technologiques (OPECST)’ hold a public hearing on Artificial Intelligence in the Senate.
Five ‘table ronde’ or round table were organised mostly with academics on the different aspects of the societal moves due to Artificial Intelligence (AI or IA in French) :
The technologies related to the AI ;
The Strategic dimensions in AI research ;
Political, societal and economical questions raised by the development of AI ;
Ethical issues raised by AI ;
Legal challenges due to the application of AI.
It was pointed that some milestone progress on deep learning has been achieved. Machines have surpassed human champions in most intellectually challenging games, including Chess, Scrabble, Othello, even Jeopardy. On March 2016, the Google AlphaGo DeepMind’s Artificial Intelligence program beat Lee Sedol, the strongest Go player in the world. Go—a 2,500-year-old game is far more complex than Chess. An exceptional powerful computer had to process more than 30 million moves. In 2011, IBM Watson won against two of Jeopardy’s greatest champions creating a big buzz in the tech community. It was early 1996 when Deep Blue won its first game against a world chess champion, when it defeated Garry Kasparov. Gerard Sabah, Research Directeur at CNRS points that IBM Watson used 3000 processors, far more than any publicly available computer, disposing over 3 million pages of encyclopaedia. Benoit Le Blanc, Lecturer in Computer Science at the Polytechnic Institute of Bordeaux (IPB) , Deputy Director of the Ecole Nationale Supérieure de Cognitique in Bordeaux (ENSC), Researcher in the Cognition and Human Engineering (CIH) team of the Laboratory of Materials Integration (IMS), Ecole Nationale Supérieure de Cognitique de Bordeaux (ENSC), insists on the importance of interdisciplinary and human-machine collaboration. He reminded that these results were only possible thanks to the feed of experienced human expert players.
Machines can process large amount of data much faster than any human. They still can’t take initiative. They can’t make sense of what they do. They would not be able to explain why they did it. Furthermore, as Francois TADDEI mentioned, machines can perform in one area, can win a specific game, remaining unable to play any other game Machines lack the capacity to question themselves, still a human exclusivity. For Jean-Daniel Kant, Professor at University Pierre et Marie Curie, the future of machine learning is in the multidisciplinary study to understand and recreate human behaviour. Collective Artificial Intelligence system can achieve multi-agents, autonomous machines. The objectif being to create intelligent machines to replace human.
David Sadek, Directeur of research at Institut des Mines, Institut des Telecoms, points that machines are not yet capable of emotional intelligence. They lack reasoning, unable of producing language or any introspection.
Jean-Gabriel Ganascia, Professor at Paris University of Science Pierre et Marie Curie, LIPSIS, President of the Ethics Committee, mentions the technical progress made in Facial recognition. Google’s Facenet accuracy is said to be of 99,63 %, higher than any human performance. The result is achieved thanks to the study of 200 million images. Another area of progress concerns Conversational agents such as Amazon’s Echo or Apple’s Siri.
Gerard Sabah explains how Google Deep Dream can recognise a cat within millions of images, however still would not know how to explain what a cat is. He sees the future of the AI in hybridation of AI and human.
Artificial Intelligence raises several ethical challenges. In case of imminence of a collision, who should the autonomous car, protect at first? The car, the driver, a pedestrian, or…? Human and machine’s logical thinking do not have same considerations. On the question of appropriate human decision making, David Sadek mentioned the a Battleship game used by US marines for training. The software always won the battle as it would its own ships drain if touched to maximise the progress.
Laurent Alexandre, Founder of Doctrissimo website and specialist of Transhumanisme, reminded there was no equivalent of Facebook or Google in Europe raising the issue of digital sovereignty.
Several speakers mentioned the challenges of sovereignty, in Political, Financial, Judicial and even personal identities in a context where social media can know more on individuals ID than the government. What is France’s position with regard to the major US companies who have considerably invested in AI- the GAFA, Google, Apple, Facebook and Amazon – financially and in human resources. on the other side, there is the BATX, Baidu, Alibaba, Tencent. Xiaomi. How could France retain its scientific forces and maybe offer more long term research budgets?
In France, academic research remains too isolated from the industry as mentioned by Yves Demazeau, Directeur de recherche au CNRS a Grenoble, Directeur de l’Association Française pour la recherche en Intelligence Artificielle (AFIA).
Gilles Dowek, Research directeur at l’INRIA and adjunct professor at l’ENS Paris-Saclay, suggested we spoke of Artificial intelligences, in plural as they are a variety of AI each with different ethical challenges.
Will machine ever be able to replace human? The impact of AI in the workplace remains uncertain. Would the least skilled jobs disappear or the intermediary level ones as consider Jean-Daniel KANT (professor a l’university des Sciences, Pierre et Marie Curie) ? How many jobs will be created by the AI ? There is no clear answer. For Marie-Claire Carrere-Gee, Présidente du Conseil d’orientation pour l’emploi, only 10% of jobs would be threatened by the AI if the market adapts itself fast enough.
Should machine have a legal personality? Alain Bensoussan is an attorney specialised in advanced technologies defining himself as “the advocate for the rights of robots and their sovereignty”. According to him, the right must adapt to the growing role that Artificial Intelligence plays in our lives.
“Human freedom is framed by the legal entity and the freedom of other holders of rights and obligations is framed by a legal person. In the same way, because robots are automatons, intelligent maR chines capable of making decisions autonomously, I think that it is necessary to create the notion of a robot person. With everything that it implies: the robot’s identification so that we know who we addressing, responsibility, a governance equivalent to that of legal persons, a capital, an insurance and full traceability, like constant monitoring of their actions to use as proof of their eventual responsibility”
Alain Bensoussan is the author of a charter on the rights of robots. See Report for EU Parliament by Mady Delvaux.
Finally, another major challenge of AI is privacy compliance. Isabelle Falque-Perrotin, Data Protection Authority at the CNIL, reminded that AI had been identified by the privacy community during the International Conference of Data Protection and Privacy Commissioners in Marakech. She pointed the four factors making AI and data protection appear antinomic : 1- AI has a massive need of data, where data protection suppose data minimisation. 2- The principle of limitation of data storage, where AI needs to store data to memorise. 3- Data security, how to secure data that is invisible? 4- data subject rights suppose an informed consent, while the AI is a machine in constant transformation.
Still many challenges to overcome for a harmonious development of artificial intelligence in accordance with human rights.
It was announced that the CNIL is launching a debate with all parties involved in AI. Following the Law for a Digital Republic, the CNIL has been tasked by Axelle Lemaire, French Deputy Minister for Innovation and Digital Affairs, to launch a wide reflection on the place of algorithms, organising a round of debate around this topic, which should be completed in September with a report on the subject.
The ‘Conseil national du numérique (CNNum)’ will have to come up with a tool to collect public’s chalenges with algorithms, while INRIA will coordinate the launch of a platform called TransAlgo, for the development of the transparency and accountability of algorithmic systems. You can watch the video of the launch of Agoranov, the new French Artificial Intelligence strategy here.
I have left out the questions of bioethics. What AI and robotics can bring to medical sciences? How will they disrupt genetics or radiology or skin cancer research for instance? Will robots replace doctors and should they become decision maker? I will post on this specific subject treated at a recent European Forum on Bioethics in Strasbourg.
Update : Google’s AlphaGo AI beats world’s best human Go player. This is a major step as Go is an even more complex game than chess responding to a more intuitive logic than the mathematical logic of the chess.
AlphaGo retires from competitive Go after defeating world number one 3-0.
Further read on Pearltrees.
This work is licensed under a Creative Commons Attribution 4.0 International License.