Talks
Prédiction des durées de séjours avec des données médico- administratives à l’aide d’un réseau de neurones
SFP#7 Artificial intelligence as Free Software
Abstract
For the seventh episode of our Software Freedom Podcast we talk with Vincent Lequertier about transparency, fairness, and accessibility as crucial criteria for artificial intelligence (AI) and why it is important for our society to release AI software under a Free Software license.
Upcoming Challenges in Artificial Intelligence Research and Development
Abstract
Artificial Intelligence is now smarter than ever, showing human-like abilities at complex tasks such as images classification or natural language processing. But despite its recent advances, it's still not a silver bullet. This talk will present a few challenges in the research and development of artificial intelligence that slow down its progress and adoption. In particular, problems around fairness, the training of models and how to share them will be introduced as well as possible Free Software solutions.
Putting Artificial Intelligence back into people's hands Toward an accessible, transparent and fair AI
Abstract
Artificial intelligence is now widespread for critical tasks such as crime recidivism risk assessment, credit risk scoring, job application review or disease detection. Because it has more and more impact on our lives, it becomes essential to make auditable AI software so that everyone can benefit from it and participate in its development. This talk will present the methods that can be used to build fairness into artificial intelligence and explain how to control its progress thanks to the four freedoms of Free Software. The talk is divided in three parts: How to make accessible Artificial Intelligences? Can AI be transparent and accurate? How to build fairness into AI? For each part the context will be presented as well as possible solutions.