Sectors, languages and software¶

In Programming you write Java. Then you open the news and every AI headline seems to mention Python. Did you pick the wrong language? Will an AI assistant write your code before you finish the course anyway?

Good questions, and they deserve data, not opinions. Today you look at three kinds of evidence: what European statistics say about which companies use AI, what real AI projects on GitHub are actually written in, and what developers themselves report in the largest surveys. Then you look at your own future job: what AI changes in software development, and what it does not.

Where AI is used most¶

Eurostat is the statistical office of the European Union. Every year it asks companies with 10 or more people which digital technologies they use.

Its 2025 figures:

So roughly one company in five in the EU uses AI. It takes root first where there are lots of digital data, repeated decisions with economic value, and staff who know how to use it: that is why IT companies and large companies lead.

Beyond the ranking, some sectors have well-known uses:

Tourism is not at the top of Eurostat's table, but it is our sector, and you already found a real Balearic case in the last lesson. Remember too that hiring, credit and health are high risk under the AI Act.

Languages and tools¶

A library is code that somebody else wrote and published so you can reuse it. Nobody writes a neural network from zero: you import PyTorch.

Around the language there are three more tools you will meet:

• Notebooks such as Jupyter or Google Colab: one document that mixes code, text and results.
• Graphics cards (GPUs), usually rented in the cloud from Unit 3, for training.
• Ready-made models and AI cloud services that you call through an API: a web address your program sends a request to and gets an answer from. Very often a company trains nothing; it consumes a model someone else trained.

In class: what are real AI projects made of?¶

In pairs. On the page of every GitHub repository, the right-hand column shows a Languages bar with the share of each language. You do not need an account.

Step 1 — Choose four repositories from this list (at least one that is not mostly Python). Write down the top three languages with their percentage, the date you looked, and what the project is for, in one sentence from its README.

Step 2 — Find and cite two figures in primary sources: (a) in Stack Overflow's 2026 Developer Survey (on stackoverflow.blog), the share of developers who use Python and the language in first place; (b) in GitHub's Octoverse 2025 report (on github.blog), the most used language on GitHub and the language that still leads AI and data science.

Step 3 — Answer: which project is "more C++ than Python", and why do you think so?

Hand in on Classroom (individual, "S18 · What are AI projects made of?", by 23:59 the same day): the table of your four repositories; the two cited survey figures; and your own paragraph (5–8 lines) on how AI would change the work of your project company's IT department. No generative AI in this activity.

AI in software development: your sector¶

Your qualification's sector is software development, and AI is changing it fast.

Look at this code. An assistant proposed it to find a user. It compiles and it works. Would you accept it?

String sql = "SELECT * FROM usuaris WHERE nom = '" + nom + "'";
Statement st = con.createStatement();
ResultSet rs = st.executeQuery(sql);

It is vulnerable to SQL injection: whatever the user types is glued into the query, so a name containing a quote can change the query itself. The correct version uses a parameterised query (PreparedStatement with ?). Generated code must be reviewed like the code of a new teammate.

SQL injection is an attack where text typed by a user is executed as part of a database query, because the program pasted it in without separating data from code.

AI will most likely not take your job away; it will change it. Whoever can only copy what a tool proposes is replaceable. Whoever can explain why that code is wrong is not.

For your project¶

This lesson feeds Deliverable 4 · AI plan. It is also the session where your group shares out the sections and roles of the deliverable:

Key takeaways¶

• About one EU company in five used AI in 2025 (Eurostat); information and communication leads, large companies use it far more than small ones.
• Python is the reference for training; C++ is often the engine; SQL prepares the data; Java, Kotlin and JavaScript integrate AI into apps.
• Many apps train nothing: they call a ready-made model through an API.
• Real projects on GitHub show it: Python on the outside, C++ on the inside. Cite the date, the numbers change.
• AI changes software development (assistants, new products, new roles) but you still review, test and answer for the code.

Quick check¶

You now know where AI is used, what it is made of and what it changes for you. Time to put it all together for your company: Deliverable 4 · AI plan.