In your career—whether it’s in business or education—mastering language and communication skills like presentations, negotiation, and job interviews is both crucial and challenging. Learners often struggle to find practice opportunities, and educators struggle to maximize interactive experiences in training sessions.
Here, the role of a GPT, or AI agent, can solve these problems.
This is why I have recently been learning about, experimenting with and making various educational GPTs, like Story-making for language learning, Negotiation trainer, and even one designed to help brainstorm ways for Gamifying your pedagogy.
What is a GPT and how is it better than an AI chatbot like ChatGPT?
A GPT is an AI agent designed to perform a specific function or set of related functions. While it uses the ChatGPT model to understand and generate human-like text based on the input it receives, the GPT is tailored to a specific function, like practicing a specific B2B negotiation, as specified by the creator.
Unlike traditional AI chatbots that require specialized prompting to set up quality functions or interactions, these GPT AI agents are already programmed to do that specific function. This allows for more focused and effective practice environments and makes GPTs exceptionally useful in language education and skill development.
Unfortunately, use of GPTs is currently restricted to ChatGPT Plus subscribers. But if you have access, the following will guide you through the creatino process.
How to make an educational GPT
After selecting the ‘Create a GPT’ option on OpenAI’s platform, you will have two options to creating the GPT: “Create” with the help or ChatGPT, or “Configure”, which means you doing the configuring.
Option 1: “Create” with ChatGPT assistance
This option involves directly chatting with the GPT to develop its functionalities. Through these interactive chats, the GPT gradually learns and adapts, and from these interactions, the “Instructions” for the GPT are formulated. This method is more dynamic and iterative, allowing the creator to guide the GPT’s learning process in real-time, shaping its responses and capabilities based on the specific requirements of the intended application, such as educational contexts.
There are pitfalls with this option because a certain amount of instruction drift will happen when you ask ChatGPT to make changes. It seems like what happens in regular chats with ChatGPT, every response is different, and so when it makes a change to the original GPT instructions, some other things may be changed. This is really frustrating. So, the better option is to “Configure” it your self.
Option 2: “Configure” the GPT independently
This option is more complex, but gives you more specific control of the GPT creation process. You have 8 steps to complete:
1. Name: Assign a unique name but clear enough for the user to know what its main function is.
2. Image: Provide an image; DALL-E will make one for you if you want, but they are often too complicated, so ask it to simplify it and make it look like a logo.
3. Description: Clearly define your GPT’s purpose and functionality.
4. Instructions: This is most important – this is where you tell the GPT to do something and how. Write clear and detailed instructions as these will act as prompts to guide the GPT’s responses.
5. Conversation Starters: You can create up to four prompt buttons that the user can click on as starting prompts.
6. Knowledge (or Training Set): Incorporate a relevant knowledge base or training set. At present, you can upload up to 20 files. Ensure to instruct the GPT to reference this material for informed responses. It appears that text data is best here, as ChatGPT has difficulty interpreting numerical data in training set analysis.
7. Capabilities: Choose functionalities like web browsing, image generation, or data analysis to assist the GPT in addressing queries more effectively. If you only want the GPT to use the Knowledge data you uploaded, then turn off these capabilities.
8. Actions and API Integration: Define specific actions your GPT can perform. This step might involve complex procedures like coding to integrate APIs for utilizing specific functionalities. Actually, the last one is more complex and requires coding knowledge or clever workarounds, as the last half hour of Liam Ottley’s GPT beginner’s guide video describes.
Now, on to some of the pitfalls I’ve experienced so far.
Pitfalls and ethical issues
When creating GPTs, especially for educational purposes, it’s important to be aware of potential pitfalls.
The first involves bias and variability in the training data. GPTs learn from the data you put in its Knowledge base, so if these data are biased, the GPT will likely inherit these biases, which can lead to inappropriate responses. But I have noticed that it seems that it will not only draw on the knowledge training set, but also use ChatGPT data. This may lead to a kind of data drift over time, away from the original knowledge base and perhaps even from the original instructions.
Related to this is Maintenance and Updating. You may need to occasionally check the GPT performance to ensure that the GPT model does not drift from the original instructions and purposes will require monitoring.
Finally, there are some potential ethical issues. Controversies about data privacy and copyright are emerging on a daily basis in this new area of commercial AI. Is the data used in the Knowledge training copyrighted? What will happen to the users’ input data? Is it sensitive? Will users need to be warned about using their own private data? These are all questions that should be considered when developing AI tools, including GPTs.
Addressing these pitfalls will require careful planning, continuous monitoring, and an ethical approach to the use of AI in education.
Monetization and broader use cases
Educational GPTs, once fine-tuned, can be integrated into training modules or even standalone language learning apps. This can open up opportunities for use outside the ChatGPT ecosystem and even the possibility for monetization through subscriptions, pay-per-use models, or by offering specialized training services. As mentioned above, there are ways to incorporate GPTs onto websites, though this requires some coding knowledge or knowledge of apps no-coding that can write the html or Java code for you.
Conclusion
GPT technology offers educators a novel approach to tackle the challenges in teaching complex skills like negotiation. By understanding its basics, creating tailored solutions, and responsibly navigating its challenges, educators can enrich the learning experience and broaden their teaching methodologies
