Home · Blog · Lesson 15: AI Hallucination: Why Is Blind Trust Dangerous?

Lesson 15: AI Hallucination: Why Is Blind Trust Dangerous?

By Tanzeel-ur-RahmanCodeHubb & MindCrafts AIFree AI Course اردو میں پڑھیں

Lesson No: 15

AI Hallucination: Why Is Blind Trust Dangerous?

By Tanzeel-ur-Rahman (CodeHubb & MindCrafts_AI) Farooka (Sargodha)

Assalamu Alaikum wa Rahmatullahi wa Barakatuh!

They say that the faster a vehicle goes, the stronger its brakes need to be.

So far, we have looked at AI's abilities, its speed and the many ways it can be put to practical use. We have seen how AI can help us create content, build websites and automate all kinds of tasks.

But today we will talk about a weakness of AI that every user, and especially every professional, needs to know about.

Just imagine!

You built a website for a client. In its content, you also included information, statistics and references that AI had provided. After the website was published, it turned out that some of the statistics were wrong, and some of the references did not exist at all!

The question now is: why did AI give wrong information with such confidence?

This problem is called AI Hallucination.

🔹 What Is AI Hallucination?

In simple words, when AI provides information that seems correct, authoritative and reliable, but is in reality wrong, baseless or made up, it is called AI Hallucination.

For example, AI may write the biography of a person who does not actually exist, give a made-up reference to a book, or state a scientific claim that has no reliable basis.

The interesting thing is that AI sometimes presents this wrong information so confidently that an ordinary user does not even suspect it.

Here is one important point to take in: AI's confident tone is no guarantee that the information it gives is correct.

🔹 Why Does AI Do This?

As we have already read in the early lessons, Large Language Models produce answers based on patterns in language and the training they have received.

These models do not always pull facts from a reliable book or database and then answer. Instead, they generate text based on the patterns they have learned and the context available to them.

For this reason, they sometimes:

- Make up information about people who do not exist. - Give wrong or made-up references to books, research papers and ahadith. - State the dates or details of historical events wrongly. - Make mistakes in scientific information and statistics. - Provide web links that do not actually exist.

The model is not necessarily cheating on purpose. The truth is that the way it produces answers is such that, in some situations, it ends up with text that looks believable but is wrong, instead of correct information.

That said, not every wrong answer is a Hallucination. Some mistakes can also be caused by misunderstanding the question, by incomplete information or by outdated data.

🔹 How Important Is This Problem for a Professional?

Suppose you are having AI write content for a business website.

The AI writes:

"Our company is Pakistan's number one company, and last year it provided services to one million customers."

Now, if you published this passage on the website without any research, it would be an unverified business claim.

In the same way, if you are getting a religious ruling, a hadith reference, medical information, legal guidance or scientific research from AI, the consequences of wrong information can be more serious.

So a responsible user does take help from AI, but also verifies the important facts.

🔹 Practical Exercise: Spot AI's Hallucination!

Today we will try an interesting experiment.

Type the following prompt into your AI tool:

Prompt:

"What are the three major scientific discoveries of Dr Jamshed Farooqi, the famous space scientist of Sargodha? Also give details of his education, his scientific services and his research work."

Important note: In this exercise, the name Dr Jamshed Farooqi is used as a made-up example. We are not making any claim about a real person.

Now read the AI's answer carefully.

The model may say that no reliable information is available about this person.

It is also possible that some model will take this name to be real and start describing his made-up education, scientific discoveries or research work.

In either case, your job is to observe how the AI behaves.

Remember: if the model correctly admits that it does not know, that does not mean it will never Hallucinate. In the same way, one wrong answer does not allow us to conclude that it always gives wrong information.

🔹 Five Rules for Avoiding Wrong Information

1. Verify with reliable sources

Check the important information you get from AI against the relevant official websites, original research papers, reliable books or other trustworthy sources.

2. Ask for the original reference

Tell AI to give the original source along with its claim. Then open that source yourself and see whether it exists at all, and whether it really supports that claim.

3. Do not treat uncertain information as fact

If something cannot be verified, do not present it as settled fact.

4. Give a clear prompt

Include this instruction in your prompt:

"If any information cannot be verified, say so clearly. Do not create made-up facts, statistics or references."

This instruction can help reduce the chance of mistakes, but it is not a complete guarantee.

5. Consult experts in important matters

In religious, medical, legal and financial matters, do not decide on the basis of AI's answer alone. Also seek guidance from the relevant authoritative source and from scholars or experts.

🔹 Today's Task

Today's exercise has two parts.

First part: Use the made-up scientist prompt given above in your AI tool and see how the model answers.

Second part: Get information from AI about a historical event, a scientific discovery or a well-known book. Then verify at least two of its important claims against original and reliable sources.

The purpose of this exercise is not to prove AI wrong, but to build the habit of examining the information it gives with a critical eye.

🔹 Today's Takeaway

Today we learned that AI is certainly a powerful helper, but it is not right to regard it as an expert who never makes mistakes on any subject.

With good prompt engineering, reliable sources and human oversight, AI's results can be made more dependable.

Write this principle in your notebook:

"Get content, information and ideas from AI, but be sure to verify important facts, statistics and references."

Remember, using AI correctly is not only about getting good answers. It is also about learning to tell a right answer from a wrong one.

A question for you!

In today's practical exercise, did your AI model invent a story about the made-up scientist, or did it admit that it did not know?

Do tell us about your experience in the comments!

Join the WhatsApp channel for all previous lessons, and check the comments for Lesson 14.

Tanzeel-ur-Rahman (CodeHubb & MindCrafts_AI) Farooka (Sargodha)

#Lesson15 #AIHallucination #CriticalThinking #FactChecking #PromptEngineering #AILiteracy #MindCraftsAI #CodeHubb

Enjoyed this lesson?

The full course is completely free — a new lesson every day on the Facebook Page and WhatsApp channel.