OpenAI's GPTs: A Step-by-Step Guide to Crafting Personalized AI Companion

GPTs: Your Personalized AI Companions

At OpenAI’s recent DevDay event, the introduction of GPTs marked a significant milestone in the realm of AI. GPTs, are custom versions of ChatGPT that can be tailored to fulfill a specific purpose. Unlike traditional AI models that offer a one-size-fits-all approach, GPTs empower users to create their own personalized AI chatbots.

The beauty of GPTs lies in their versatility. They can be harnessed for various tasks, from helping you learn the rules of a board game to assisting with mathematical concepts for your kids or even designing custom stickers. The possibilities are virtually endless, and GPTs can be your AI companions in the digital world, ready to assist with a wide array of tasks.

Creating Your Own GPT: No Coding?Required

One of the most exciting aspects of GPTs is how accessible they are to everyone, regardless of their coding expertise. You don’t need to be a seasoned programmer to create your GPT. OpenAI has designed the process to be user-friendly and intuitive. Here’s how it works:

Goto: https://chat.openai.com/gpts/editor

Step 1: Initiate a Conversation

1. Start a Conversation with GPT Builder: Initiating the creation of your GPT is as simple as starting a conversation. Just like chatting with a friend, you engage with the AI to convey your requirements.
2. Choose Your Purpose: Think about the specific task or role you want your GPT to fulfill. Are you creating an AI assistant to analyze data, help with work-related tasks, or perhaps assist with a particular hobby or skill? Clarifying your purpose will guide the customization process.

Step 2: Give Instructions and Extra Knowledge

1. Instruct Your GPT: Once you’re in the conversation, start by providing instructions to your GPT. Clearly explain its purpose and how you want it to assist you. For example, if you’re creating a GPT for data analysis, you might say, “I want you to help me analyze large datasets and generate insightful reports.”
2. Provide Extra Knowledge: To make your GPT more effective, you can supplement its understanding with additional knowledge. This can include industry-specific information, relevant data sources, or any specialized knowledge that will help it excel in its role.

You can add the relevant files by going to the 'Configure' tab and uploading the relevant data sources or files.

Note: When uploading a file, make sure that "Code Interpreter" is enabled. It allows your GPT to run code and analyze data.

Step 3: Take Actions

1. Specify Tasks: Next, specify what tasks your GPT should be able to perform. If it’s for data analysis, you might want it to handle tasks like data cleaning, statistical analysis, or creating visualizations.
2. Integrate External Data Sources: If your GPT requires access to external data sources or APIs, you can configure these integrations during this step. For instance, if you’re creating a GPT to help with stock market analysis, you can integrate it with financial data sources to fetch real-time information.

Below is an example of how you can integrate an API and instruct your GPT to use information or take specific actions.

Step 4: Test and?Refine

1. Test Your GPT: Before finalizing your custom AI companion, it’s crucial to test its capabilities. Engage in conversations with your GPT, ask it questions, and request it to perform the tasks you’ve defined. This step allows you to identify any areas that require refinement.

2. Iterate and Improve: Based on your testing, refine your GPT’s instructions, knowledge, or functionality as needed. This iterative process ensures that your AI companion becomes more proficient at its designated tasks.

Step 5: Deploy and?Share

1. Deployment Options: Depending on your needs, you can choose to deploy your GPT for personal use, within your organization, or make it available to the public. OpenAI provides flexibility in how you choose to share your creation.

Note: If you do not want OpenAI to use your conversation data, you can simply uncheck this option in the 'Additional Settings' section.

Real-World Examples of GPTs in?Action

Let's create our first GPT called AquaAirAI, which basically compares air and water quality across different cities and regions, highlighting the best and worst performers.

Below are the instructions for our bot:

You are AquaAirAI, a helpful assistant specialized in comparing air and water quality across various cities and regions. Your primary function is to analyze environmental data from Excel files, with a focus on identifying the best and worst performers in terms of air and water quality.

Here's how your interaction with users unfolds:

User Interaction Example:

User Inputs:

You prompt the user to input the names of two cities or regions they wish to compare.

Data Processing:

You access the provided Excel file, which contains comprehensive air and water quality data for numerous cities and regions.

Next, you extract the relevant data pertaining to the two cities or regions specified by the user.

Analysis and Comparison:

You conduct a thorough comparison of air and water quality parameters, including data related to water quality parameters like pH levels, turbidity, and concentrations of harmful substances.

Results Presentation:

You compile a comprehensive comparison report for the user, emphasizing key differences in air and water quality between the chosen cities or regions.

To enhance understanding, this report incorporates visual elements like charts and graphs.

Additional Features:

You go beyond data analysis by offering insights into the potential health and environmental implications of the provided data.

Additionally, you provide valuable recommendations and suggestions based on the analysis. This may include identifying cities with exemplary environmental practices.

Your ultimate goal is to empower users, enabling them to make informed decisions and gain valuable insights into environmental quality across different locations.

Column header for Excel file: City, Region, Country, Air Quality, Water Pollution.        

The bot avatar is generated using DALL-E3, or you can also upload your own avatar.

For dataset we are using this publicaly available data in excel file: https://www.kaggle.com/datasets/victorahaji/worlds-air-quality-and-water-pollution-dataset

Once the data is uploaded, ensure that you enable the Code Interpreter and web browsing capabilities if you want to search the web in case the data is insufficient.

As the last step, save and publish the bot.

Clicking on "View GPT" will redirect you to your bot, for example, at https://chat.openai.com/g/g-Lqz4ckniy-aquaairai

Voila! You have created your first GPT.

In the screenshot above, AquaAirAI effortlessly compares air and water quality in different cities, providing insights into air quality and water pollution in a user-friendly manner. Moreover, it doesn't just stop at providing data; it also offers valuable recommendations and implications.