- Artificial intelligence (AI): An attempt to reproduce cognitive skills (such as computer vision, speech synthesis and recognition, knowledge representation, and machine learning…) that are unique to human beings – to support decision-making and automation.
- Generative AI (GenAI): A subset of AI relying on deep learning techniques, targeting content creation. Using text (or voice) prompts (as inputs), it generates responses in the form of text, images, audio, music, video, code, etc.
- Machine Learning: This is a field of AI development to give machines the ability to “learn” from data, via mathematical models. Three main approaches are used: supervised learning, unsupervised learning, and reinforcement learning.
- Deep Learning: This is a machine learning process using (artificial) neural networks with several layers of hidden neurons. These algorithms use a large number of parameters, and they require a very large amount of data to be trained.
- Hallucinations: Generative AIs can in certain (relatively rare) cases, give “delusional” answers with no link to reality or truth. They can “invent” characters, events, dates, etc. that never existed.
- Bias: An anomaly in the responses of machine learning algorithms, generally based on three different sources: data acquisition, user interaction, and/or amplification from the predictive model.
- Prompt: The data input or question asked to a generative AI tool, in the form of a text question, with or without additional documents.
- Algorithm: A finite, unambiguous sequence of instructions and operations to solve a class of problems.