Today, OpenAI has announced the launch of GPTs on its blog, which are custom versions of its ChatGPT generative AI that allows anyone to create versions of the AI that are more suited to specific tasks to fit the ways that users want to use it.
Custom Instructions have been available from OpenAI in ChatGPT since July, which allows users to set some preferences, but users have been asking for more control and the ability to further adjust the AI ever since, which this launch aims to satisfy.
OpenAI has also announced that users can also share GPTs publicly, with the launch of the GPT Store later this month allowing verified builders to publish and share creations with the rest of the world. Additionally, OpenAI will spotlight specific creations that it thinks are the most "useful and delightful" in different categories such as productivity, education, and "just for fun."
OpenAI goes on to emphasize the privacy and safety controls within GPTs, and chats within them are not shared with the builders. If a GPT uses third-party APIs, a user can choose whether or not to share their data with that API. Additionally, it has set up new systems internally to help review GPTs against its usage policies to prevent users from sharing harmful GPTs.
As mentioned above, developers have the ability to link GPTs to external APIs and other plugins, such as databases, emails, or e-commerce platforms. The possibilities for this are wide and far-reaching, further expanding ChatGPT"s influence on the world.
Finally, OpenAI has announced that customers of OpenAI Enterprise have the ability to deploy GPTs exclusively on an internal basis, being securely published within workplaces. ChatGPT Enterprise users have the ability to start creating GPTs from 8th November, controlling how GPTs are shared and whether external GPTs can be used inside the business, and chats aren"t used to develop OpenAI models.