There are two ways to create your own customized GPT. One is to use the wizard inside the website, which requires no coding and makes it easy to upload your own data and custom instructions. Those GPTs are then only available to other OpenAI premium users. Alternatively, you can code to the AI Assistants API, which is more complex, but more flexible, and allows you to incorporate AI capabilities into your systems directly.

Making a custom GPT can be easy, but it's still an AI. That means even with your own data supplied, it will tend to favor information in its training set. Getting an AI to do what you want (and only what you want) can be difficult.

This talk covers both approaches: making a custom GPT inside the web site, and coding your own. It talks about both the costs and benefits of either approach, and, unlike other tutorials available, the coding part uses Java rather than Python or TypeScript.