Jean-Gabriel Ganascia

Jean-Gabriel Ganascia — latest news, discussion & facts — IJR

Research

In a 1991 paper on deriving learning bias from rule properties, Ganascia introduced a measure of rule interestingness. This measure is now known in the data mining literature as the Ganascia index. It is used for scoring partial classification rules by weighing supporting records against counter-examples, and is also known as descriptive confirmed-confidence.

In the same 1991 paper, Ganascia described CHARADE, which is a system for learning sets of classification rules. He developed it as part of his work on symbolic machine learning. Johannes Fürnkranz and Tomáš Kliegr wrote that the system simplifies its output in two stages. First, it removes conditions from individual rules that are logically implied by the remaining conditions, and then removes whole rules that are implied by other rules in the learned set. Fürnkranz and Kliegr credit it as the first rule-learning system to recognize the trade-off between a rule set's accuracy and its comprehensibility.

With Julien Bourdaillet, Ganascia developed MEDITE, which is a monolingual text aligner built for textual genetic criticism. It compares the successive versions of an author's text. It detects insertions, deletions, replacements and displacements at the level of individual characters, using an approach related to sequence alignment in bioinformatics. With Pierre Glaudes and Andrea Del Lungo he developed Phoebus, a system for detecting quotations and reuses in literary texts. It indexes fixed-length word sequences in both texts before repeatedly identifying and extending matching passages.

Ganascia's work in computational ethics takes a top-down, logic-based approach. He has translated ethical principles into formal representations a machine can reason with. In 2007 he formalised Aristotelian consequentialism, Kantian ethics, and Benjamin Constant's theory of principles, using non-monotonic logic implemented in answer set programming. With Fiona Berreby and Gauvain Bourgne he modelled the principle of double effect. Later, he used the event calculus to represent ethical scenarios under both consequentialist and deontological frameworks.

Ganascia and Thomas Powers wrote a chapter on the ethics of AI ethics to The Oxford Handbook of Ethics of AI (2020), which identified conceptual ambiguities in the language of AI ethics and distinguished an AI agent, which has agency but not intention, from the ethical conception of an agent that intends its actions upon reflection.

Le Mythe de la Singularité

In 2017 Ganascia published Le Mythe de la Singularité: faut-il craindre l'intelligence artificielle?. It is an essay against the technological singularity. In the book, he challenges three propositions on which the idea rests: Moore's law, Ray Kurzweil's law of universal evolution, and the prediction that machines will become autonomous. He writes that it is incorrect to assumed that the exponential growth in machine performance will continue indefinitely and that computing power does not amount to intelligence. He also writes that, in a philosophical sense, reinforced learning does not give machines autonomy.

He characterises the singularity as belonging to mythos rather than logos. He compares its structure to ancient gnostic thought and argues that its proponents are led by the idea of uploading the mind into a dualism inconsistent with their own premises. In the concluding chapter he attributes the promotion of the idea by major technology companies to corporate communication strategy, and argues that it diverts attention from their ambition to assume functions historically held by states.

Reviewing Le Mythe de la Singularité in RIMHE, Philippe Monin called the book a courageous intervention. He wrote that few voices had questioned a view shared by figures such as Stephen Hawking and Elon Musk. Monin found the chapter on modern gnoses the most stimulating in the book. Brigitte Munier, in Hermès, La Revue, described it as courageous and necessary. She wrote that Ganascia had produced a clear argument which can be understood by readers outside the field. She noted as distinctive among books on the subject his decision to concentrate on the scientific status of the critical point rather than on the wider catalogue of posthumanist promises. She found the essay hurried, however, and regretted that he did not develop his closing argument further. In Études, Franck Damour wrote that the book deconstructs a myth and examines what drives it. He described the critique of the reasoning underlying the myth as sober and clear. Writing in the Swiss daily Le Temps, the philosopher Mark Hunyadi called the attention that the book draws to myths artificial intelligences generates as its chief merit. However, he wrote that a complete account would also have to ask what makes audiences credulous.

Jean-Gabriel Ganascia

Jean-Gabriel Ganascia

Article

Related journals

Related journals

Article

Early life and education

Career

Research

Writing

Voir et pouvoir

Le Mythe de la Singularité

Later books

Awards and honours

Selected publications