Where does AI get its information?
Objective: This worksheet addresses the key question of how artificial intelligence (AI) acquires its knowledge and how reliable this information is.
Content and methods: The material conveys that AI systems are not conscious beings, but are based on calculations derived from training data (e.g. books, articles, websites) and external sources (e.g., databases, news). The worksheet uses true/false tasks, text work, mind map creation for source overview, and practical case analyses to critically reflect on the functioning and reliability of AI.
Competencies:
- Media criticism: Recognizing that AI responses are based on probabilities and may be incorrect.
- Information research: Understanding the need to compare information from different sources (AI vs. books).
- Data protection awareness: Raising awareness of how user data is handled when interacting with AI.
Target group and level: Middle School
50 other teachers use this template
Target group and level
Middle School
Subjects
Where does AI get its information?
Introduction
Artificial intelligence (AI) is part of our everyday lives, for example in voice assistants, translation programs, and chatbots. But how does AI actually know so much?
Look at the picture and answer the questions.
Right or wrong?
Decide and justify whether the statements are true or false.
🤖 Statements about artificial intelligence
Check the box and explain your decision:
| Statement | Correct | Incorrect | Reason |
|---|---|---|---|
| AI reads books independently, just like a human being. | |||
| AI is trained with large amounts of data. | |||
| AI can have its own feelings. | |||
| AI always selects the most truthful information from the internet. | |||
| AI can reflect the same prejudices found in human-written texts. | |||
| AI recognizes patterns in data. |
Assignment
Read the information text and select the correct statement in each case.
Where Does AI "Chat-GPT" Get Its Information?
The artificial intelligence "Chat-GPT" can answer questions, write texts, and provide insights. But where does it actually get its information from? Firstly, Chat-GPT was trained on a vast amount of data. This includes texts, articles, books, and web pages. Through this data, the AI learns to recognize patterns and connections in language. It doesn’t understand content like a human but calculates probabilities to generate suitable responses.
For queries that require current information, Chat-GPT can access external sources. These may include online databases, news websites, or information systems. This allows the AI to provide answers that are up-to-date, even if the data was created after the initial training period.
When a user asks a question, Chat-GPT processes it using its training data and available external sources. The queries are not stored permanently; they are used only to calculate the quickest and most relevant answer. This ensures a level of data protection and security for users.
Despite its capabilities, Chat-GPT has limitations. It doesn’t possess consciousness, cannot think or feel, and doesn’t form opinions. All of its knowledge is based solely on the data it was trained with and the external sources it accesses.
Mind map
Create a mind map in which you clearly present the most important information from the text.
1. Write "AI – Where does the information come from?" in the center of your mind map.
2. Create at least three main branches.
3. Add 2–3 keywords or short explanations from the text to each main branch.
Use:
• Keywords instead of whole sentences.
• Arrows, colors, or symbols to make connections visible.
Lade Zeichenfeld...
Discussion cards
The class works in small groups (2–4 people). Each group receives cards with AI scenarios.
1. Each person receives their own card with an AI scenario.
2. First, read your card quietly to yourself.
3. Then, each person briefly presents their scenario to the group.
4. Finally, discuss the questions together as a group.
⏱️ Time for group discussion: 15 minutes.
Important:
• Each person must express at least one opinion.
• Answers should refer to the information text.
Results are presented briefly in a plenary session or voted on by a show of hands.
Scenario:
Scenario:
Scenario:
Scenario:
Write down your collected opinions on the scenarios here.
Reflection
Answer the following questions.
The “Western Bias”
Much of the training data on the internet comes from Europe and North America.
Discuss in your group:
- What impact does this have on AI responses when asked about the traditions, values, or history of countries in the Global South?
Give a specific example of a possible biased perspective.
What impact does this have on AI responses when asked about the traditions, values, or history of countries in the Global South?
Solution for teachers
🤖 Statements about artificial intelligence – sample solution
| Statement | Correct | Incorrect | Reasoning |
|---|---|---|---|
| AI reads books independently, just like a human being. | ✔ | AI processes texts technically, but does not understand them like a human being. | |
| AI is trained with large amounts of data. | ✔ | Machine learning is based on large training data sets. | |
| AI can have its own feelings. | ✔ | AI simulates emotions, but does not feel them. | |
| AI always selects the most truthful information from the internet. | ✔ | AI selects information based on algorithms and data, not truthfulness. | |
| AI can reflect the same prejudices found in human-written texts. | ✔ | AI can inherit biases present in training data. | |
| AI recognizes patterns in data. | ✔ | Recognizing patterns is a core function of AI. |