With her latest book, Laurence Devillers establishes herself as one of France’s leading experts in automation, robotics and AI. In it, she undertakes a useful exercise in analysing the lessons to be be learned from the exponential development of generative AI since 2022. This exercise is becoming increasingly difficult due to the recent proliferation of books, articles and talks, which are more or less “hallucinatory”, – on this “disruptive innovation” and this historic “social phenomenon”. The author clearly explains how to “survive with AI”, by harnessing its digital capabilities without losing its human capabilities, that is to say, unlike software, to retain one’s ability to understand situations, to make sense of phenomena and to become aware of one’s responsibilities.
She endeavours to convince the reader that, although the the development of AI is inevitable, it can nevertheless be controlled by humans, thanks to practices governed by rigorous ethical charters and to appropriate actions for learning its systems, its logic and its languages. She criticises approaches that ascribe to AI a consciousness and a personality. She identifies the multiple biases of its software but puts into perspective the risks that AI poses to the psychology of its users and to democracy.
She answers 8 of the questions that her users ask themselves: How can deepfakes be recognised? What is a ChatGPT hallucination? from ChatGPT? What are the hyperparameters of ChatGPT? How do you train ChatGPT to perform a task and make it repeat it? What are the limitations of current AI? What language does AI speak to us in? What is the paradox of our use of Big Tech? How can we foster critical thinking in young children when it comes to digital technology and AI? The author also endeavours to shed light on the 8 myths surrounding AI. Finally, she explores the scientific potential of AI, its contribution to public health, health, its environmental impact and its contribution to the graphic arts.
Laurence Devillers is a professor at Sorbonne University and director of the LISN laboratory at the CNRS.
Jean-Jacques Pluchart