Last Updated: April 2023
As artificial intelligence (AI) becomes increasingly integrated into businesses and organizations, board directors are increasingly expected to have a basic level of understanding of AI algorithms in order to perform an effective role in corporate governance.
This is because AI can have significant impacts on a company's operations, risk management, and strategic decision-making. Board directors who are knowledgeable about AI algorithms are better equipped to assess the risks and opportunities associated with AI initiatives, ensure that AI systems are used ethically and responsibly, and make informed decisions about investments in AI.
Furthermore, understanding AI can help board directors ask the right questions, evaluate the quality of AI-related proposals and reports, and communicate effectively with other stakeholders. Overall, having a basic understanding of AI algorithms is critical for board directors to fulfill their oversight responsibilities and make well-informed decisions in today's data-driven business environment.
Before we start let’s just take a minute to get some jargon out of the way. The word “Features” is very commonly used when discussing how AI works and it refers to something very specific. If you don’t know what it means specifically in this context, please read this article first:
Now that we have that out of the way, let’s get straight into it. AI algorithms are typically separated into three groups: Supervised, Unsupervised and Reinforcement Learning. Understanding this distinction provides a framework for organising these algorithms in your mind.
4 Broad Types of Machine Learning / AI Algorithms
- Supervised learning: This type of machine learning involves training a model on labeled data, where each example in the data has a corresponding label or target variable. The goal is to learn a function that can accurately predict the label for new, unseen examples.
- Unsupervised learning: In unsupervised learning, the model is trained on unlabeled data and must discover patterns and structure in the data on its own, without the guidance of a labeled target variable. The goal is often to identify groups or clusters of similar examples or to learn a compressed representation of the data.
- Semi-supervised learning: This is a combination of supervised and unsupervised learning, where some of the data is labeled and some is unlabeled. The goal is to learn a function that can accurately predict the label for new, unseen examples, while also leveraging the additional information provided by the unlabeled data.
- Reinforcement learning: In reinforcement learning, the algorithm learns by interacting with an environment and receiving feedback in the form of rewards or penalties. The goal is to learn a policy that maximizes the cumulative reward over time.
Top Supervised Learning Algorithms
- Classification: A machine learning algorithm that is used to predict the class or category of a given input. For example, given a dataset of pictures of animals labeled as "cat" or "dog", a classification algorithm can be trained to predict whether a new picture is a cat or a dog based on certain features. Classifiers can also be applied to text. For example, determining whether an email is spam or not based on its content.
- Regression: A machine learning algorithm used to predict a continuous value or a numerical output. For example, predicting the price of a house based on its features such as location, size, and number of bedrooms.
- Decision Trees: A machine learning algorithm that uses a tree-like structure to model decisions and their possible consequences. Each branch in the tree represents a decision, and each leaf node represents a possible outcome. For example, a decision tree can be used to determine whether a customer will buy a product based on their demographic information and purchasing history.
- Convolutional Neural Networks: A type of neural network commonly used in image recognition and processing. It uses a process called convolution (sorry to use convolution to explain convolution but it really takes more time to explain) to identify and extract features from images, making it possible to classify images based on those features.
- Large Language Models / Transformer-based Neural Networks: are primarily supervised learning models. They are trained on vast amounts of labeled data, such as text documents with pre-existing labels or tags, to learn patterns and relationships between words, phrases, and concepts. During training, the model is presented with inputs and their corresponding outputs, and it adjusts its internal parameters to minimize the difference between its predicted output and the actual output. This process of adjusting parameters continues until the model can accurately predict the output for new inputs. However, LLMs can also incorporate unsupervised learning techniques in certain stages of their training. For example, some LLMs use unsupervised pre-training, where they are first trained on a large dataset of unlabeled text data to learn general language patterns before being fine-tuned on smaller, labeled datasets for specific tasks.
Top Unsupervised Learning Algorithms
- Bayesian Networks: A probabilistic graphical model that represents a set of random variables and their conditional dependencies via a directed acyclic graph. It is commonly used for reasoning under uncertainty, such as in medical diagnosis or financial forecasting.
- Sentiment Analysis: A type of text classification algorithm that determines the emotional tone or sentiment of a piece of text, such as a product review or a social media post.
- Named Entity Recognition: A natural language processing technique used to identify and extract specific entities from text, such as names of people, places, organizations, and products.
- Rule-Based Systems: An AI algorithm that uses a set of predefined rules to make decisions or perform actions. For example, a rule-based system can be used to diagnose faults in a machine based on a set of rules that describe the behavior of the machine.
Note that some of these algorithms can be used for both supervised and unsupervised learning, depending on the task at hand. For example, decision trees and neural networks can be used for both classification (supervised) and clustering (unsupervised) tasks.
Currently, supervised learning algorithms are used more frequently in corporate life than unsupervised learning - though this could evolve as the field continues to advance.