I therefore challenged ChatGPT on the calculation’s accuracy:
This again underlines the importance of not taking ChatGPT’s word as being perfect, and of challenging it.
Here’s the transcript of this discussion in PDF format.
Activate GPT-4 (Recommended but Not Obligatory)
To get the most out of ChatGPT’s ability to solve problems, I’d recommend using the GPT-4 algorithm, which is more powerful than the default engine (GPT-3.5).
If you don’t have these options when you start a new conversation, I’ll cover how to set up ChatGPT Plus in a later chapter dedicated to this subject.
Let’s take a look at some additional examples for solving problems with ChatGPT.
Ask ChatGPT a Problem-Solving Question
We can start by asking some simple math questions:
🧔🏻♂️: What is the annual water consumption of a household consisting of four people who are at home for an average of 200 days per year and who use an average of 25 gallons of water per day?
🤖:
To calculate the annual water consumption for a household, you can use the following formula:
Annual Water Consumption = Number of People × Average Daily Water Use per Person × Number of Days at Home per Year
Given the values you provided:
Annual Water Consumption = 4 people × 25 gallons/person/day × 200 days/year
Annual Water Consumption = 20,000 gallons/year
Therefore, the annual water consumption for the household is 20,000 gallons per year.
As you can see, ChatGPT is more than capable of working this out.
Try to get into the habit of asking ChatGPT to help solve complex problems for you. Remember to always be cautious about its responses and don’t be afraid to correct it when it gets things wrong so you can get better results.
You can actually guide ChatGPT in its reasoning using “chain of thought” prompting.
Boost ChatGPT Using “Chain of Thought Reasoning”
There are a few ways of doing this:
Basic: just add the instruction “Provide step-by-step details of your reasoning” to your question and this will help to improve results on its own. For example:
“What is the annual water consumption of a household consisting of four people who are at home for an average of 200 days per year and who use an average of 25 gallons of water per day? Provide step-by-step details of your reasoning.”Reasoning by example: give an initial question-example and the answer-example you expect to get with details of the reasoning used. Then ask a second (different) question and ask ChatGPT to respond. It will model its response on your method.
🧔🏻♂️:
QUESTION: Roger has five tennis balls. He buys two additional bags of tennis balls. Each bag contains three tennis balls. How many tennis balls does he have now?
RESPONSE: Roger has five balls to start with. Two bags, each containing three tennis balls, makes six tennis balls, so 5 + 6 = 11. The answer is 11.
QUESTION: There are 23 apples in the canteen. If 20 apples are used to make lunch and six additional apples are purchased, how many apples remain?
🤖: To find out how many apples remain, you can follow these steps:
Start with the initial number of apples: 23.
Subtract the number of apples used to make lunch: 23 - 20 = 3.
Add the number of additional apples purchased: 3 + 6 = 9.
Therefore, there are 9 apples remaining in the canteen.
Step-by-step reasoning: in this case, we’re not giving an example, but we’re providing a logical structure that we want ChatGPT to follow, step by step. It will respond using the steps you’ve provided and this will often help you to get a better quality result.
🧔🏻♂️: You work for a sales team generating challenges to motivate the team to achieve their sales targets. You must use the following format to create the message announcing the next challenge to the team:
Step 1: choose a theme, which might be linked to the season (Halloween, Christmas, etc.), to a sporting event (Olympic Games) or something else.
Step 2: choose a prize that is linked to this theme, which will be awarded to the person with the best sales results.
Step 3: write a message announcing the challenge to the team.
I hope that these examples will inspire you in your conversation with ChatGPT. As is often the case, you need to have the confidence to guide it so that its thought process works the way we want it to. I think you’ll be pleasantly surprised by the results! 🤩
Let’s Recap
GPT-4 is capable of handling complex problems and explaining its reasoning to provide better responses compared to GPT-3.5.
Adapt your questions to help ChatGPT work through problems logically, but you should always check and correct the response if required.
Use “chain of thought reasoning” to improve the quality of its responses, asking for a logical explanation at each step of its reasoning, giving a detailed example or guiding ChatGPT through each step.
This is the end of the second part of the course. Now it’s time to review what you’ve learned so far. Let’s go!
