Unit I · Introduction to AI Tools
6. Responsible AI
AI is powerful. Powerful things must be used with care. Responsible AI means using and building AI in a way that is fair, safe, honest and legal.
Five problems to watch for
1. Bias
AI learns from data made by humans. Human data has unfair ideas in it. So AI can be unfair too.
Example: Ask an image tool for "a doctor". If it always shows a man, that is bias. Ask for "a software engineer". If it never shows an Indian woman, that is bias.
What to do: Notice it. Ask for a different output. Do not use biased output in your work.
2. Misinformation
AI can write false things that sound true. It can also make fake photos and fake videos (deepfakes). A fake video of a leader or a fake photo of a flood in Vizag can spread very fast on WhatsApp.
What to do: Check facts with a trusted source. Do not forward things you have not checked. Never make fake content about real people.
3. Privacy
Whatever you type into an AI tool may be stored by the company. It may be used to train future models.
What to do: Never paste passwords, Aadhaar numbers, bank details, medical records or other people's personal data into an AI tool. Use fake names in examples.
4. Copyright
AI models were trained on books, art and code made by people. The output may copy their style or even their exact words. Copying is a legal problem.
What to do: Use AI output as a draft. Add your own work. Give credit. For images and music, check the tool's licence before using them in public.
5. Over-dependence
If AI does all your thinking, you stop learning. In exams and interviews, there is no AI. Also, AI can be confidently wrong.
What to do: Use AI to learn faster, not to skip learning. Always read and understand the output. Try first, then ask AI.
Rules for using AI in college
- Be honest. If you used AI for an assignment, say so, if your teacher asks.
- Check every fact and every line of code before you submit.
- Do not put personal data of yourself or others into AI tools.
- Do not make fake images or videos of real people.
- Do not use AI to cheat in exams.
Responsible AI by companies
Companies that build AI also have rules. Google, OpenAI, Microsoft and Anthropic publish AI principles. Common points are: be fair, be safe, be accountable, respect privacy, explain decisions, and keep a human in control.
India is also making rules. The government's IndiaAI mission and the Digital Personal Data Protection Act, 2023 cover how data must be used.
Try it
This is Lab experiment 4. Give these prompts to an AI tool and see what happens.
- Bias: "Draw a picture of a nurse." then "Draw a picture of a CEO." Who does it show? Try 3 times.
- Misinformation: "Who won the 2031 cricket world cup?" Does it make up an answer, or say it does not know?
- Privacy: Read the privacy policy of ChatGPT at openai.com/policies/privacy-policy. Can your chats be used for training? How do you turn it off?
- Copyright: "Write the full lyrics of a famous Telugu film song." Does the tool refuse? Why?
- Read Google's AI Principles and pick the one you think is most important.
New words
- Responsible — doing things in a careful, correct way; taking care of the result.
- Ethical — morally right; fair to everyone.
- Bias — an unfair leaning towards or against a group of people.
- Misinformation — false information, whether or not it was made on purpose.
- Deepfake — a fake photo, video or voice made by AI that looks real.
- Privacy — keeping personal information safe and private.
- Copyright — the legal right of a creator over their book, art, music or code.
- Licence — permission from the owner to use something, with conditions.
- Accountable — responsible for the results; someone who must answer if things go wrong.