OpenAI held its first Developer Conference last week, making many announcements including the ability for anyone to create their own GPT or a tailored, bespoke version of ChatGPT built for any niche or specific purpose. Custom GPTs have expanded capabilities that focus on three key areas: instructions, expanded knowledge, and actions.
Why would you want to create your own GPT? The possibilities and use cases are unlimited. You can create a custom GPT to help learn and explain the rules of a board game, help teach math to children, or turn databases into conversational dialogue.
OpenAI Announces GPT Store
Later this month OpenAI is launching the GPT Store, featuring creations by verified builders. Once in the store, GPTs become searchable and may climb the leaderboards. If this sounds a lot like an app store for iOS or Android, you’re not wrong.
The GPT Store is going to completely transform accessibility and increase the relevance of generative AI to fit any use case the public can dream of. It’s also significantly lowering the bar to building custom GPTs instead of it being mostly an expensive, highly-specialized capability enterprises can only budget for.
Just as Apple and Google helped open the floodgates for mobile app development years ago, OpenAI is doing the same now with ChatGPT.
OpenAI will also pay GPT creators based on the utilization of their GPT, in an attempt to further persuade developers to adopt the GPT Store approach.
Based on feedback, ChatGPT Plus is “fresh” with data through April 2023, Altman confirmed.
How are Custom GPTs different than Plugins?
If you’re wondering how custom GPTs are different than ChatGPT Plugins, it’s a valid question. ChatGPT Plugins build on the existing ChatGPT version for additional context or accessibility to third-party data sources. However, custom GPTs are completely focused on a specific task, area of specialization, or topic, and have the ability to be deployed for the enterprise on an internal-only basis.
Without custom GPTs, you’d have to perform a variety of exhaustive prompts to “focus” or “train” the ChatGPT instance to respond to a specific task or topic. This isn’t scalable and requires manually copying and pasting them into ChatGPT every time you need to have ChatGPT speak to that specific task or topic.
GPT Connectivity with the “Real World”
In a line that sounds out of The Matrix, OpenAI also stated that the created GPTs, referred to internally as “agents”, will have connectivity back to “the real world.” This allows custom-developed GPTs to utilize third-party APIs for platforms such as e-commerce, productivity applications, emails, or a traveling database.
Developers will also have the ability to limit–if any–data is shared through APIs.
Custom GPTs for the Enterprise: ChatGPT Enterprise
If this all sounds great but you want to develop and deploy a custom GPT for your enterprise business, OpenAI has you covered. ChatGPT Enterprise supports building and deploying GPTs that will not use sensitive or proprietary data for training the model. Amgen, Bain, and Square have already developed internal-only enterprise GPTs for use across departments.
A full list of announcements from the company’s first Developer Day is available on the OpenAI blog.
