This morning my 11 years old son came to me with a math question from one of his practice books. After trying out few minutes, we could not get a convincing answer. So, referred the answer page of the book and understood how it is supposed to be resolved. At that moment I thought of why can’t try out with intelligent AI chatbots out there to understand the steps. Later only to realize how useless these generative AI stuffs are at least for solving logical math problems.

## Here is a Question from Book

Below is the math question from a book titled “Challenging Maths Primary 5”.

There were 176 more people in Hall A than in Hall B. When 435 people moved from Hall A to Hall B, there were 3 times as many people in Hall B than in Hall A. How many people were there in Hall B at first?

### Answer from ChatGPT Supported App

First, I asked the question to one of the ChatGPT supported apps downloaded from Apple iPhone App Store. The app told me the answer was 611 people initially in hall B.

### Answer from Google Bard

Next, I headed to Google Bard which is using the latest Gemini AI model. There are three draft answers and the first one is -777 and Bard declared the information in the question is insufficient as number of people can’t be in negative.

Second answer is 46.5 which is also ruled out by Bard itself as the answer can’t be in fraction. Even it says, the answer does not make any sense.

Third draft shows the answer is 652.5 and Bard rounded it to 652 as number of humans can’t be in fraction. I don’t know how it make sense to round of now which can’t be done in previous draft version.

As you can see in the left sidebar, Bard gives a creative title to the question as “Hall B occupancy puzzle” on its own. However, the answers are completely wrong with no logical sense.

### Answer from Bing AI Chatbot

Finally, here is what Bing AI in Windows 11 Copilot tells and the answer is there are 476 people in hall B. It can’t be the answer as adding 435 people from hall A will result in 911 which will have reminder when divided by 3 (as it should be 3 times of hall A people).

After one more try, Bing AI chatbot told me 346 people in hall B which is completely different than previous answer (476). Though 346 looks correct, it is not as (346+435)/3 will result in a fractional number again.

I went ahead and told Bing AI that the correct answer is 606 and it did not accept it at all. It was so absurd saying, “1041 is not equal to 3 times 347. You can verify this by multiplying 347 by 3 and comparing it to 1041”.

It went ahead with its funny explanation of, “if 1041 is equal to three times of 347, then it should also be true that 347 is equal to three times of 1041”.

### Answer from Book

Here is the logical explanation from book with the answer as 606 people are in hall B which means there are 782 people in hall A (606 + 176). This is the correct answer as 606+435=1041 which is the number of people in hall B at the end. This 1041 is exactly 3 times of people in hall A at the end as 782-435=347. There are no negative or fractional numbers in this answer.

As mentioned, at least it is easy to match the answer with logics given in the question and come up with no negative or fraction numbers. I did ask one of the math teachers and initially she could not give correct answer. But once I shared the book answer, she also matched it with the question and accepted that’s the correct answer.

## Conclusion

Annoying part with these AI chatbots is that they give different answers each time when the book has one simple answer. For example, Bard gives three drafts which also changes next time. Same thing with ChatGPT apps and Bing AI that the answer changes drastically. The worst part is that none of these AI bots gave the book answer even a single time. Bing is so aggressive to convince you that if 3*347=1041 then 1041*3=347 with useless explanation.

The problem I can see is these AI bots are good at scrapping content from various websites and vomiting with paraphrasing them. But they are poor in logical expressions with factual content something like the question above and never accepts the factual inputs because they are not trained for that. Though math teachers can also make mistakes, they are logical and rely on the facts. So, if you want your kid to think and solve math problems logically, find a math teacher (or teach yourself) and do not look for an AI bot.

## Leave a Reply

## Your email is safe with us.