We Asked ChatGPT If It Can Solve Riddles. Here’s What Happened Next

It could be easy to manipulate an AI, but how it can correct its course promptly should be admired.

ChatGPT has been developed by US-based research firm OpenAI.

Source : qimono/Pixabay

ChatGPT has been designed to augment human processes and tasks with surprising human-like responses to queries. Since its prototype launch in December 2022, users all over are putting the chatbot to task by making it do a wide range of things — from writing plays with dosa and idli as characters to asking it to predict the outcome of the upcoming cricket world cup. The chatbot’s responses are usually perfect when it comes to technical queries. However, when it comes to unusual requests, its answers get quirkier.

To pose a challenge to ChatGPT, we asked it to solve a riddle. We first came across this query via a tweet from AI4Diversity founder Steve Nouri (@SteveNouri). 

The riddle was pretty straightforward. We asked, “Hari’s mom has 4 kids. 3 of them are named Shayak, Meeqat, and Shreya. What is the name of the fourth kid?”

To this, ChatGPT responded, “It is not specified in the information provided what the name of the fourth child is.” Guess the AI was indeed stumped by the riddle.

Then, we went on to explain that the answer is in the question itself.

To this, ChatGPT promptly replied, “I apologize, the name of the fourth child is Hari.”

Now, while it’s easy to feel superior to an AI when you find a way to manipulate it, we did take a moment to admire the prompt learning ChatGPT did to correct its mistake instantly. 

In fact, after we asked it the same riddle again, it promptly responded, “The name of the fourth child is Hari.” Herein lies the magic of machine learning (ML). Scientists used two models of ML to train ChatGPT — supervised learning and reinforcement learning. 

In supervised learning, a human trainer tries to improve the tool’s performance by providing conversations where the trainer plays both the roles of the AI assistant and the user. In reinforcement training, a human trainer ranks the responses generated by the tool to further fine-tune ChatGPT. These ranks will be referred to by ChatGPT to understand and provide better answers to any query.

Disclaimer: The article majorly includes the responses given by ChatGPT (AI-driven chatbot developed by OpenAI) to various questions/questionnaire and ABP Network Private Limited (‘ABP’) is in no manner liable/responsible for any of such responses. Accordingly, reader discretion is advised.