Unit 4 · Artificial intelligence¶
In a city-builder game such as Cities: Skylines, there comes a point where you stop changing every traffic light by hand. You switch on an automatic advisor: it watches the traffic overlay, learns when the jams happen and tunes the lights by itself. It is not magic and it is not always right. It is only as good as the data it watches, it costs money to run, and somebody still has to check what it does. That advisor is artificial intelligence. In this unit you train one in your browser, test a generative AI critically, calculate whether an AI project pays for itself, and find out what AI is really made of, and what it changes in your own future job.
| Learning outcome | Hours | Term |
|---|---|---|
| RA4 — Identifies applications of artificial intelligence (AI) in environments of the sector the qualification belongs to, describing the improvements its implementation brings. | 5 | 2nd |
What you will learn
- Explain what AI, machine learning and deep learning are, and avoid the trap of calling everything automatic "AI"
- Train an image classifier in the browser and explain why it fails
- Use a generative AI critically: check its answers against sources and spot hallucinations
- Classify AI uses by the risk levels of the European AI Act
- Find real AI uses in Balearic companies and estimate, in a spreadsheet, whether an AI project is profitable
- Name the languages and tools AI is built with, and describe how AI is changing software development
flowchart LR
D["<b>Data</b><br>(the traffic overlay)"] --> M["<b>Model</b><br>(the automatic advisor)"]
M --> DEC["<b>Decision</b><br>(tune the lights)"]
DEC --> R["<b>Result</b><br>(fewer jams, lower cost)"]
H["<b>People</b><br>(the mayor checks)"] -.-> DECThe model is the part everybody talks about. This unit is just as much about the other boxes: where the data come from, who checks the decision, and whether the result is worth the cost.
Lessons¶
| # | Lesson | What it covers |
|---|---|---|
| 1 | What is AI | AI, machine learning and deep learning; training and inference; your first model with Teachable Machine |
| 2 | Generative AI and responsible use | Large language models, hallucinations, rules for responsible use and the risk levels of the AI Act |
| 3 | Automation and profitability | Rules vs AI, automating vs optimising, a real Balearic case, and a profitability spreadsheet |
| 4 | Sectors, languages and software | Which sectors use AI most (Eurostat), the languages behind real AI projects on GitHub, AI in software development |
| 5 | Deliverable 4 · AI plan | What you hand in, who writes what, how it is assessed, the defence |
What you are graded on¶
The six assessment criteria of the unit's learning outcome, the lessons that train each one, and the part of Deliverable 4 where you show it.
| Criterion | Official wording, in English | Lessons | Deliverable 4 |
|---|---|---|---|
| a) | The importance of AI in the automation and optimisation of processes has been identified. | Automation and profitability · What is AI | Section A |
| b) | AI has been related to the massive collection of data (big data) and its processing (analysis) with the profitability of companies. | Automation and profitability | Section B |
| c) | The present and future importance of AI has been assessed. | What is AI · Generative AI and responsible use | Section C and group part (risk levels, rules) |
| d) | The sectors with the most relevant implementation of AI have been identified. | Sectors, languages and software · Automation and profitability | Section C |
| e) | The programming languages used in AI have been identified. | Sectors, languages and software · What is AI | Section D |
| f) | How AI influences the sector of the qualification has been described. | Sectors, languages and software | Section C |
There is no exam. The unit is assessed through Deliverable 4 (70%: your own section, your defence and the group part) and the class activities handed in on Classroom in sessions 15 to 18 (30%). The rules are on the course project page.
AI in an AI unit
The module's rule still applies here: generative AI is not allowed in assessed work unless the teacher authorises it for a specific activity. In this unit exactly one activity is authorised: the error-hunting activity in session 16, in class. Everything else, including Deliverable 4, is written by you.