Board directors need high-level knowledge about AI and machine learning models to effectively evaluate and decide upon AI strategies.
Therefore, the resourceful and pro-active chairman of my client invited me to run a workshop together with the rest of the board.
It was an efficient way for them to gain some relevant knowledge – just prior to the kick off of their AI strategy process in collaboration with their internal experts.
We covered quite a lot in an afternoon!
1. The Basics:
Introduction to the basic types of machine learning models such as supervised learning (e.g., linear regression, decision trees), un-supervised learning (e.g., clustering, dimensionality reduction), and reinforcement learning.
Each type has its own use cases, strengths and weaknesses.
2. Purpose of the Model:
Understand what a model is designed to achieve.
Is it for prediction (e.g., sales forecasting), classification (e.g., spam detection), clustering (e.g., customer segmentation), or something else? And why it matters?
3. Performance Metrics:
Query the key performance indicators (KPIs) used to evaluate the model – why this is relevant.
We briefly touched upon: – accuracy, precision, recall, F1 score, ROC AUC for classification models; Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) for regression models; and silhouette score, Davies-Bouldin index for clustering models. I detected some enthusiasm….
4. Data Requirements:
Different models require different types and different amounts of data.
An initial understanding of these requirements can provide an insight into the suitability of a model for a given task.
5. Compliance and Ethics:
What are some important aspects to consider to ensure the models comply with relevant regulations and ethical guidelines, especially when dealing with sensitive data.
6. Model Interpretability:
Some models, like decision trees, are easier to interpret than others, for example neural networks.
Depending on the application, the ability to understand and explain a model’s decisions may be important.
7. Maintenance and Update Requirements:
Models may need to be retrained or updated as new data becomes available.
Understanding what these different requirements are, can help assess the ongoing resources needed.