Who makes and programs artificial intelligence?
Behind every AI system there are many people – not just programmers. In this lesson you will learn who takes part in making artificial intelligence, who the biggest model “makers” are today and where people work on artificial intelligence in Serbia.
This page describes the current situation (November 2026) and is updated once a year.
How an AI system is made
- Idea and problem: what the system should do (for example recognize plant diseases from a photo).
- Data: examples are collected and labeled (thousands of photos labeled with the disease). This is often the biggest job.
- Training: programmers choose the type of model and let computers learn from the data – for large models, thousands of special computers work for days or months.
- Testing: the model is tested on examples it hasn't seen; people look for mistakes and unfairness.
- Use: the model is built into an app, website or device; how it works is monitored and it is improved.
Who works on artificial intelligence
| Profession | What it does |
|---|---|
| researcher (scientist) | comes up with new ways of learning and new types of models; most often a mathematician or computer scientist |
| machine learning engineer | trains and tests models, prepares data |
| programmer | makes apps and websites that use models |
| data engineer | collects, cleans and stores huge amounts of data |
| data labeler | labels examples (for example “there is a pedestrian in this image”) |
| safety and ethics expert | checks that the model isn't dangerous, unfair or misused |
| designer | designs how people use an AI tool |
| domain expert | a doctor, farmer, teacher… – explains the problem and checks whether the results make sense |
The good news: to work with artificial intelligence it isn't enough just to know programming – math, language, logical thinking and knowledge from other fields are important too.
Who makes the best-known models today
The biggest models, like those in chatbots, are made by companies and research centers that have a lot of money, experts and computers. Among the best known are:
| Who | Well-known products / models |
|---|---|
| OpenAI (USA) | ChatGPT |
| Google / Google DeepMind (USA, United Kingdom) | Gemini, AlphaGo, AlphaFold |
| Anthropic (USA) | Claude |
| Microsoft (USA) | Copilot |
| Meta (USA) | Llama (open models) |
| Mistral AI (France) | Mistral (partly open models) |
| companies from China (for example DeepSeek, Alibaba) | DeepSeek, Qwen |
- Closed models are used over the internet (an app or website), and the “inside” of the model isn't public.
- Open models can be downloaded and run on your own computer; that is why universities and small teams all over the world also work on them.
- Besides large models, there is a huge number of small models for one job – they are also made by individuals, school students and university students. You will make your own in this course too.
Artificial intelligence in Serbia
- The Institute for Artificial Intelligence of Serbia (based in Novi Sad) does research and projects applying AI, for example in agriculture, medicine and flood protection.
- The Government of Serbia has adopted the Strategy for the Development of Artificial Intelligence for 2025–2030.
- At universities (electrical engineering, mathematics, computer science) there are courses and study programs in artificial intelligence and data science, and many local IT companies develop AI solutions.
How to start right now
- Learn programming – the most used language is Python (SR).
- Don't neglect math: probability, charts and functions are the “language” of machine learning.
- Make a small project – in the module How a machine learns you will make a model without a single line of code.
- Follow competitions and workshops in computer science; more and more of them have tasks about artificial intelligence.
Let each student choose one profession from the table and describe in 3 sentences what their working day would look like on the project “AI recognizes plant diseases”.
Common misconceptions
- “AI makes itself” – people choose the data and the way of learning, and they test and improve the models; without people there is no artificial intelligence.
- “Only programmers work on AI” – experts in data, safety, design and the field where AI is used are also needed.
Practice tasks
Try it yourself first, and only then open the solution.
- Put the steps of making an AI system in order: training, use, data, testing, idea.
- What is the difference between an open and a closed model?
- Why is labeling data an important job?
Solution 1: the order
idea → data → training → testing → use (and after that monitoring and improving).
Solution 2: answer
An open model can be downloaded and run on your own computer (and studied). A closed model is used only through the company's service, and its inside isn't public.
Solution 3: answer
Because the model learns from labeled examples. If the labels are wrong or incomplete, the model will learn the wrong things – “garbage in, garbage out”.
Check your knowledge
Open the quiz
Quiz: who makes artificial intelligence
1. Which step is often the biggest job when making an AI system?
2. What is an open model?
3. Which programming language is used the most in artificial intelligence?
4. Why is a model tested on examples it hasn't seen?