Comment tirer le meilleur parti du machine learning, l’apprentissage machine, dans les 5/10 ans à venir? Le rapport de la Royal Society britannique, publié le 25 avril dernier, explore en profondeur les enjeux et les implications du développement du machine learning,
Qu’est ce que le machine learning ? : “Machine learning is a form of artificial intelligence that allows computer systems to learn from examples, data, and experience. Through enabling computers to perform specific tasks intelligently, machine learning systems can carry out complex processes by learning from data, rather than following pre-programmed rules” est-il expliqué dans le rapport. Les données étant devenues les carburants du machine learning, émergent de nombreuses incertitudes sur le caractère privé de certaines de ces données, ainsi que sur la portée réelle du consentement des individus à leur utilisation.
Le rapport identifie les questions suivantes comme les priorités pour la recherche dans le domaine du machine learning :
• “Interpretability: can we create powerful machine learning systems where the reasons for particular decisions or recommendations can be understood or interrogated?
• Verification and validation: can we create more advanced, and accurate, methods of verifying machine learning systems?
• Privacy: what are the technical solutions that can maintain the privacy of datasets, while allowing them to be used in new ways?
• Fairness and dealing with real-world data: how can real-world data be curated into usable forms, addressing ‘real-world’ messiness, and systemic – or social – bias?
• Causality: How can machine learning methods discover cause-effect relationships?
• Human-machine interaction: How do we design machine learning systems so they can work with humans safely and effectively?
• Security: How do we ensure machine learning systems are not vulnerable to cyber-attack?”
0 commentaire