Welcome to AI Explained: From Intuition to Understanding
This series began in one of my classrooms.
While teaching Neural Networks and Support Vector Machines, I realized that what I thought was a polished set of slides (on which I had spent many weeks) was not quite doing the job I had hoped for. My students were still uncertain about some of the ideas I was trying to explain. And it was almost immediately clear that the problem was not with my students. Nope! The problem was with me and my rather abstract approach to an already abstract topic.
The painful realization reminded me of a recurring experience with AI. For the past few years, whenever I have told people about my background in AI, one of their first questions has been whether I personally believe that AI is going to “destroy humanity”. I will write about this aspect of AI in the coming weeks and months, but for now, suffice it to say that the short answer I always give is “No, it will not!”.
On the other hand, it had become rather clear to me that my bland explanation of some key AI topics to my bright and attentive students was getting very close to destroying their hope of understanding the most critical concepts in the field. And that made me wonder: Is it the prospect of AI destroying humanity, or the way AI experts explain their knowledge and vision, that creates such a frightening image?
Anyway, as a result of the ensuing self-reflection (a week from hell, which ended in a much-needed “aha” moment), I realized that I needed another way to help those ideas make sense to my students. And so, I changed the setting. That is how Fernanda, an estate agent, and her colleagues Omar and Ricardo came into the picture.
Instead of jumping directly into a technical setting (which made perfect sense to me, but less so to my audience), I started with everyday questions:
- How might Fernanda decide whether a house belongs in a particular price category?
- How could she and her colleagues work together to estimate its price?
- And when their estimate is wrong, how would they decide what to change?
These situations gave us a way to explore inputs, weights, biases and learning, with small calculations we could follow together using a simple spreadsheet or pen and paper.
AI Explained will extend this approach across 24 topics. Fernanda will remain our recurring companion, sometimes at work, sometimes with friends, family or neighbors. Each situation will help us explore an important AI concept, moving from an intuitive explanation to the details of how it works.
I will add articles over the coming weeks as time allows. We will use stories, clear explanations, worked examples and occasional exercises, and discuss where our human comparisons reach their limits.
The Journey Ahead: 24 Topics
To give you a better idea of what AI Explained will cover, here is the planned 24-topic roadmap. As each article is published, its title below will become a link to the full explanation.
| No. | Main Article Title | Subtitle |
|---|---|---|
| 1 | Should Fernanda Follow a Rule or Learn from Experience? | Understanding Rules and Machine Learning |
| 2 | How Does Fernanda Put a Price on a Home? | Understanding Linear Regression |
| 3 | How Does an Estate Agent Make a Yes-or-No Decision? | Understanding the Threshold Logic Unit (TLU) |
| 4 | What Changes When Three Agents Work Together? | Understanding the Multilayer Perceptron (MLP) |
| 5 | The Estimate Was Wrong. What Should the Team Change? | Understanding Loss, Gradients and Backpropagation |
| 6 | Has Fernanda Learned or Just Memorised the Houses? | Understanding Overfitting and Generalisation |
| 7 | When Is a Wrong Answer More Costly Than Another? | Understanding Evaluation Metrics and Decision Thresholds |
| 8 | Is Fernanda Making Her Method Too Complicated? | Understanding Regularisation |
| 9 | Which Question Should Fernanda Ask First? | Understanding Decision Trees |
| 10 | Can Several Imperfect Advisers Make a Better Decision? | Understanding Ensembles: Bagging and Boosting |
| 11 | Where Should Fernanda Draw the Dividing Line? | Understanding Support Vector Machines (SVMs) |
| 12 | What If No Straight Line Can Separate the Cases? | Understanding Features and the Kernel Trick |
| 13 | Can Fernanda Discover Groups Nobody Named? | Understanding Clustering |
| 14 | How Can Fernanda Make a Map of Similarity? | Understanding Embeddings |
| 15 | How Does Fernanda Find a Pattern in a Picture? | Understanding Convolutional Neural Networks (CNNs) |
| 16 | What Must Fernanda Remember as the Story Unfolds? | Understanding Recurrent Networks (RNNs), LSTMs and GRUs |
| 17 | Whose Words Matter for Understanding This Sentence? | Understanding Attention and Transformers |
| 18 | How Does Fernanda Continue an Unfinished Message? | Understanding Generative Language Models |
| 19 | Should Fernanda Answer from Memory or Check the Records? | Understanding Retrieval-Augmented Generation (RAG) |
| 20 | How Can Fernanda Learn When Nobody Tells Her the Best Move? | Understanding Reinforcement Learning |
| 21 | How Does Fernanda Turn a Goal into a Sequence of Actions? | Understanding Planning and Tool-Using AI Agents |
| 22 | Did the Renovation Raise the Price or Was Something Else Happening? | Understanding Correlation, Causation and Interventions |
| 23 | How Sure Should Fernanda Be? | Understanding Uncertainty and Calibration |
| 24 | What Should Fernanda Delegate and What Should People Decide? | Understanding Human–AI System Design and Oversight |
If you have ever followed an explanation of AI and still wondered, “But what is actually happening?”, this series is for you (and that is all the marketing I am consciously doing!). Oh, and please register to receive updates. Freely share the link with others who may find the topics interesting and, more importantly, give me feedback 😊.
