ChatGPT Plugins

An overview of the newly introduced plugins for ChatGPT

Posted by vahid on March 26, 2023

OpenAI has recently introduced plugin capabilities for ChatGPT, along with a few sample plugins that integrate with services such as Kayak, Shopify, and Zapier. This development provides a great opportunity for developers to start creating their own plugins, it can be a category for developers to jump in and start developing plugins like the time they developed mobile apps. It also presents an opportunity to take User Interfaces to the next level.

What are they?

Currently, OpenAI has introduced a few plugins, in addition to third-party plugins. One of these is a browser plugin powered by Bing search, which provides citations and allows users to follow up with the source of the information provided. Another plugin is a code interpreter that can convert natural language queries into Python code and display the results.

On their blog, they stated:

‘Language models today, while useful for a variety of tasks, are still limited. The only information they can learn from is their training data. This information can be out-of-date and is one-size fits all across applications. Furthermore, the only thing language models can do out-of-the-box is emit text. This text can contain useful instructions, but to actually follow these instructions you need another process.’

Also they mentioned:

‘[Plugins give the language models] access to information that is too recent, too personal, or too specific to be included in the training data.’

How to create one?

It’s described in the documentation, in overall you need to :

-        Create a manifest file

-        Develop the plugin to handle API calls

-        Register the plugin with ChatGPT

-        Activate the plugin in a ChatGPT user account

In a conversation with the user, a compact description of the plugin gets injected into the conversation, which is invisible to users. The model may invoke an API call from the plugin if relevant, and the API results are incorporated into its response. Also, The model may include links from API calls in its response, displaying them as rich previews using the OpenGraph protocol.

Conclusion:

Could it be possible to create an operating system-like platform that enables developers to create extensions or applications? From a business standpoint, involving others in the platform could be the right move to benefit both the developers and the service. This is similar to the model used by operating systems that are often available for free, and on which developers build an ecosystem of applications.

I don't think of ChatGPT as a portal to everywhere on the internet. Instead, it's more like a service that underlies many places. For example, you wouldn't necessarily go to ChatGPT to order something from Amazon; rather, you'd go to the Amazon website and use a ChatGPT-powered assistant to find what you're looking for. Something similar to the Cloudflare services that many websites use now a days. It's an underlying service that powers those websites and solves a specific problem. As we're familiar with, almost all websites have a search bar. It's likely that these search bars will evolve to become more chat-like and use natural language to help users find the material they're looking for.

آموزش مقدماتی پایتون 10 آموزش مقدماتی پایتون 9 آموزش مقدماتی پایتون 8