Generative artificial intelligence This is kind of Artificial intelligence (Artificial Intelligence) is capable of creating content such as text, images, or even videos, based on data analysis and learning.
Machine learning For AI, it is the process of learning from a database. To exploit this data, machine learning can use different methods: linear regression, decision trees, neural networks, etc.
Deep learning Subclass of Machine learning Deep learning uses neural network as its architecture. It is particularly powerful for performing unsupervised learning but requires a large database and thus high computing power.
Supervised learning The fact that AI relies on data explicitly identified by humans. For example, after ingesting thousands of images of cats identified by humans, AI will be able to detect similarities and recognize the cat. In contrast, unsupervised learning relies on data that has not been labeled by humans.
Algorithm It is a series of logical instructions to achieve a goal. An algorithm is a diagram written to be understood by humans (computer scientists, mathematicians, researchers, etc.), but translated concretely into a computer program.
Neural network It is a computer architecture in information technology where each computing unit processes a very specific function. A computational neuron fires when it specifically identifies the information it is processing. Facial recognition uses a neural network where each neuron processes data: nose, eyes, skin, etc.
Chat bot A computer program capable of speaking to a human in writing or voice.
Master's A “big language model”, which translates to “big language model” in French, is an artificial intelligence program that has absorbed a large enough amount of data to be able to interact with humans. The GPT model (developed by OpenAI) of the Chat GPT bot is the most popular of all as compared to other competitors like twin (Google) or Claude (anthropic) .
Urgent engineering It is the art of knowing how to communicate effectively with artificial intelligence to make the most of its capabilities. This is concretely the text written by the user who is formulating the request. It's about asking the right questions to get the relevant answers.
fine tuning This consists of using an artificial intelligence model to run programs that will be specialized in a specific procedure. Many startups are fine-tuning: they are creating and marketing their software based on already existing AI models.
Strong AI/Singularity The AI systems currently being developed fall into the category of “weak AI,” which is supervised by humans. The concept of “strong artificial intelligence” is still in the science fiction stage. This would be the emergence of an artificial intelligence which, through learning, would end up being close to the intelligence of the human mind and could lead to an event called the Singularity, i.e. the AI gaining “a form of consciousness.”
