Introduction:
On November 6th, OpenAI hosted its first-ever developer conference called DevDay in San Francisco, where it announced a platform for building custom chatbots.
How to use it:
Creating a custom ChatBot is very easy. First, you should have ChatGPT Plus, then click the ‘Explore’ button.
Next, click on ‘Create a GPT.’
Then, you have two options for creating a custom ChatBot: ‘Create’ and ‘Configure’. This article will focus on the ‘Configure’ option. We will create a functional chatbot for a management course.
In the ‘Configure’ tab, there are many options: Name, Description, Instructions, Conversation Starters, Knowledge, Capabilities, Actions, and Additional Settings.
The ‘Name’ will determine how our ChatBot will be named. In this example: ‘Management Assistant at KU ChatBot.’
The ‘Description’ is self-explanatory. It helps users understand the purpose of the chatbot. For example:
‘Instructions’ is the box where the prompt for the chatbot’s specialization is placed. Prompts for chatbots can be very versatile, from a personal therapist to the CEO of a big company. For this article, the prompt is as follows:
‘Conversation Starters’ are helpful tools for users who are unsure what to ask the chatbot. They are useful for common uses of chatbots like a shop assistant or as the first bridge between a customer and a company. For example:
Moving on to ‘Capabilities’, you can choose what the chatbot can access. Currently, there are three options:
- ‘Web Browsing’: allows the chatbot to access the internet. Helpful for gaining insights from open sources.
- ‘DALL·E Image Generation’: allows the chatbot to generate images using DALL-E 2. Useful in almost all creative work.
- ‘Code Interpreter’: allows the chatbot to interpret and run code, read files, analyze data, and much more. Note that other functions of chatbot creation are locked by this function. Useful in automation, code debugging, and much more. For example:
Returning to ‘Knowledge’, this gives the ChatBot context. Context is embedded information that provides the chatbot with knowledge from PDFs or other text types. In this example, a PDF provided by professors from KU will be used. If the ChatBot’s ‘Code Interpreter’ functionality is not enabled, this won’t work.
‘Actions’ allow the ChatBot to use pre-written functions. This topic is too advanced for this article, so we will skip this option.
‘Additional Settings’ have only one checkbox. If checked, it allows OpenAI to use the created ChatBot to improve this new option.
Conclusion
I hope this introduction was helpful and allowed some people to understand how to create a ChatBot using OpenAI.
Sources:
youtube.com/watch?v=EWdCMPnm8uY
theverge.com/2023/11/6/23948619/openai-chatgpt-devday-developer-conference-news
AI:
Bing AI