From the course: ISO/IEC 42001:2023: Understanding and Implementing the Artificial Intelligence Management System (AIMS) Standard

Understanding the categories of AI uses

From the course: ISO/IEC 42001:2023: Understanding and Implementing the Artificial Intelligence Management System (AIMS) Standard

Start my 1-month free trial Buy for my team

Understanding the categories of AI uses

“

- [Narrator] Imagine AI as a large diverse family with each member having unique traits and ability. The star of the family currently is generative AI. McKinsey research indicates that generative AI applications stand to add up to $4.4 trillion to the global economy annually. However, this course, the ISO IEC 42001, Artificial Intelligence Management System standard covers the entire family. For instance, robotics involves designing, constructing, operating, and using robots to perform tasks in the physical world. Robotics is a practical application of AI. On the other hand, natural language processing deals with the interactions between computers and humans through everyday language, enabling AI to understand and respond to human communication. Computer vision focuses on enabling computers to interpret and understand the visual world, a crucial capability for AI to interact with the physical environment like driverless cars. Machine learning is a significant capability or subfield in the AI family that contains generative AI as a subset. Machine learning allows the computer to learn without explicit programming. Two sub capabilities of machine learning are supervised and unsupervised models. Supervised learning models have labels or tags like name, type, or number. This allows a system to sort elements into like sets. The results are expected or otherwise produce an error. Unsupervised learning models look at raw data to see if it can be categorized into previously unnoticed groups. Deep learning is a significant subset of machine learning. It uses artificial neural networks inspired by the human brain. It operates on labeled and unlabeled or raw data, most of which are unlabeled. Labeled data helps to focus the task while unlabeled data helps to generalize to produce new examples. Generative and descriptive are two types of learning models. A discriminative model is a type that is used to classify or predict labels for data sets. Discriminative models are typically trained on data sets of labeled data points, and they learn the relationship between the features of the data points and the label. For instance, training a system that all these are candy! Once a discriminative model is trained, it can predict the label for new data points. A generative model generates new data instances based on a learned probability distribution of the existing data it is trained on. Generative is typically trained in unsupervised or semi-supervised context. This means the generative model can create graphic art and pictures, video form, music, computer code, and literary works like a new book! A helpful analogy is that a generative model simulates a human artist. They can create something based on their impression or understanding of the world. Discrimative models are like art critics. They don't create anything new, but classify and typify existing art. So for instance, if a generative model created something new, a discriminative model can categorize and distinguish it accurately. While creating new written pictorial and audio works dominates the market, it doesn't cover the entire capabilities of what needs to be managed in the AI family. Awareness of these other capabilities and categories creates the understanding necessary for a comprehensive management. As we consider establishing an AI management system, keep in mind that the requirements of the management systems are agnostic and support any member of the AI family that our business uses.

Contents