Big Data and Artificial Intelligence – A framework of responsibilities to be defined
In collaboration with Maître Nicolas HERZOG, associate-founder of the Cabinet H2O Avocats
Artificial Intelligence (IA), for a long time relegated to the rank of science-fiction fantasy, finds today multiple concrete applications linked to Big Data. Such evolutions are made possible by the massive digitalization of human activities.
Combination of Big Data and Artificial Intelligence
Medicine is the perfect example of the association of Big Data / AI allowing today to consider if not automatic diagnosis, the provision of decision support systems. In mammography, for example, « pattern » recognition software highlights suspect zones on the clichés, then checked by radiologists. Another example: the laboratory of Artificial Intelligence in Stanford created a wide database of images of skin cancer and created its algorithm to identify a potential risk. Results were highly accurate, the machine obtained results similar to the conclusions of 21 dermatologists.
Big Data and Artificial Intelligence allow as well to schedule a new deployment way of the tools of cybersecurity. As Edward Snowden said in its own time, the NSA developed a platform called MonsterMind that is not only capable of applying a protection strategy, but also to automatically fight back an attack: a modern time « War Game ».
The Open Law
Closer to our insurance and loss adjustment professions, Law is also directly concerned. Maître Nicolas HERZOG – Avocat au Barreau de Paris – Cabinet H2O Avocats – developed this subject in an article particularly enlightening “Big Data and Open Law – from the opening of judicial data towards the advent of tools of predictive justice” (Dalloz Avocats – n°1 – Janvier 2017).
It is on the basis of the article 15 of the Declaration of Human Rights and of the citizen according to which « the society has the right to ask to any public agent to report about its administration » and the evolution of a legislation from 1978 to 2016 (law Lemaire of 7th October 2016) that it was stated about « the free provision to the public and in an open format of the exhaustivity of judgements pronounced by all the jurisdictions of the judicial and administrative order and freely reusable by automatized data treatment». Maître HERZOG reminds that there are more than 4 000 000 justice decisions given each year in France.
It is on this basis of this colossal core of data that 5 French start-ups created an offer « legatech » of predictive justice. On the basis of keywords seized in a search engine, intelligent algorithms scan all the decisions of justice to help chose the best strategy allowing to refine the evaluation of risks.
New applications, new risks
The great advantage and the great danger of these tools constituted by the Big Data / IA is the fact that they are self-learning. They tend to become more and more appropriate as their knowledge base grows.
The examples we talked about here above are not trivial: they concern directly human individuals or companies’ health.
And what if a dispute arises? Actually, a failure in the execution of a performance will lead to the research of causes and responsibilities.
In the case, how responsibilities will set between data quality, algorithm power and human intervention? Or, all or a party of these tools will become autonomous in a long term? Without falling into a post-apocalypse caricature, seizing the destiny of the World by robots (Terminator, Matrix…), this option does not seem crazy.
There is no doubt that an in-depth discussion about human intervention is absolutely needed. Actually, be it about law, health or any other field, nobody doubts that mankind, judicial or technical council will allow to keep ethics and discernment in the use of the tool.
Within the limits of current technologies, this measure seems to be about common sense for learning systems do not become self-powered systems nourishing themselves of their own decisions, that will then reproduce to the infinity. The perfect example is without any doubt the system « evidence-based sentencing » employed in some American States, in charge of calculating the duration of imprisonment of prisoners to reduce the risk of relapse. The Obama administration dropped it because they found out that it gave discriminatory results.