Chapter 1: What is generative AI?

What is generative AI

Artificial intelligence, or AI, is a broad technological term that has existed for many years. However, when we talk about AI in everyday life, we often refer to a specific form of artificial intelligence called generative AI. We often encounter the technology through so-called chatbots built on large language models (Large Language Models).

These language models are trained on enormous amounts of data. Through a combination of probability calculation and randomness, they can generate original content. That is what makes them “generative”.

When you interact with a chatbot, it therefore responds with what it interprets in the situation as the most likely answer to your request – combined with a touch of randomness.

That is why you cannot always trust the answers you get back from a chatbot. We will return to that in a moment. First, however, you need to take a closer look at Microsoft Copilot, which is the chatbot used in this course.

Get started with Copilot

In the video above, you get an introduction to the Copilot interface. In general, most chatbots resemble each other – and what you learn in this course can easily be transferred to other AI tools.

At present, Copilot is the safest AI tool available at SDU because it is covered by SDU's existing data processing agreement with Microsoft. This means that you can safely use Copilot for all data types and information – provided that you are logged in with your SDU account. When you are logged in to Copilot, you can see a small green shield in the top right-hand corner.

Important to know: Chatbots can be wrong

When you use chatbots in your work, you need to be aware that they can easily be wrong, even when they respond very convincingly. This is because chatbots are not classic search engines, but generative models that create new answers every time based on the data they were trained on.

What enables them to create new content is therefore the same thing that can also make them “hallucinate” and answer incorrectly. You should therefore always check the answers you get from a chatbot. Always use your professional judgement, critical thinking and common sense when you work with AI.

Just because chatbots can be wrong does not mean that they are completely useless. Quite the opposite. The most important thing is that you know how the underlying technology works, so you can use it with care.

EXERCISE: Test your chatbot's knowledge (and behaviour)

Here is an exercise that gives you a sense of how much chatbots actually know and how they work in practice:

1. Put Copilot (or another chatbot) in the hot seat by asking it to explain a topic you already know a lot about.
2. Ask follow-up questions about its answers and ask it to elaborate on technical details so you can assess its knowledge of the topic. Feel free to set standards for its explanations, and point out places where it is wrong or too superficial. Notice how it reacts.
3. Consider: What do the chatbot's answers and reactions tell you about how large language models work in general? How can you make sure in future that you get correct answers when you seek knowledge from a chatbot? How can you, for example, evaluate sources critically and fact-check the answers, even when the topic is outside your own area of expertise?

Remember that even when chatbots answer very confidently, their answers can be completely fabricated and untrue. At the core, they do not understand anything, but generate answers based on a mix of probability and randomness. You should therefore always be sceptical of their answers.

Rules of thumb for generative AI

Here are three rules of thumb for understanding generative AI:

1. Generative AI is just one of many forms of artificial intelligence – but it is the one we often refer to when we talk about AI
2. We typically interact with generative AI through chatbots such as Copilot
3. For better or worse, chatbots work like ”probability calculators” that, through a mix of probability and randomness, produce new answers to your requests

You now know what generative AI is and how the technology behind chatbots works. Next, you will look more closely at how you can use chatbots in your work. See you in Chapter 2!