I recently read a blog about a certain writer using ChatGPT as their go-to information finder. I know there are loads of us who do that because it makes life a lot easier.
However, all Generative AI tools come with a disclaimer (the tiny letters at the bottom of the document to be read using a microscope ๐): Crosscheck the information it is offering because Generative AI can make mistakes.
What they don’t tell you is that some of it can amount to outright lying!
There is a simple reason behind it. Generative AI is not a search engine. It is a tool designed to sound like a person who knows stuff. You ask it a question, and it tries its best to answer. Unfortunately, if it doesn’t have the answer, it still tries. Sometimes when it doesn’t know its stuff at all, it lies through its virtual teeth, gives false information and false citations.
If you ever saw “Peep and the Big Wide World” on You Tube, its character Quack can give you a gist of the way Gen AI thinks. Here’s a link to one of my favourite videos: Link
I am saying this because we have seen AI behaving like Quack at work, several times over. For example, we were searching for information for a high-end leadership training we were creating. So, obviously I took AI’s help. Who doesn’t? I searched for 10 techniques leaders can use to enhance a particular skillset. I had also named a set of reputable websites to search the information from. The idea was to filter down the techniques and zero down on 3-4 most effective ones. As a response to the query, I received a long list of techniques along with citations from various reputable resources.
The knee-jerk response is to add everything given by AI. But we have been trained at office not to trust AI blindly.
So, I opened the citations and a couple of techniques were not there in the said citation. I tried other links and they weren’t there at all. All the so-called “researches” that “supported them” had never happened and the statistics were forged. It was a lie–plain and simple–to increase the number of techniques to 10 as I had asked for. Had I used the data, it could have landed me into a lot of trouble with the client.
What I am trying to say is that, if you are asking AI how many cat-crazy women are there in each country, understand that there is a high chance you will get an answer with detailed statistics, and it would be a complete lie because nobody has asked me yet how many cats I own. ๐คฃ
Having said that, Generative AI is right around 80 to 90 percent of times. It is okay to use it. With caution. If you are searching for something important, double-check all citations it is offering.
- Does the cited book exists?
- Is it written by the same expert or backed by the same organisation as AI claims?
- Did they actually say anything on your subject?
- Did they say exactly what AI is saying or is AI interpreting the views too openly?
In essence, check the engine oil, fuel, water and anti-freeze before entering the car and flooring the accelerator. ๐


