Tool Information
ChatGPT is an AI tool implemented with initial support for plugins. These plugins are designed specifically for language models with safety as a core principle and enable ChatGPT to access the most up-to-date information, run computations, and use third-party services. The tool benefits from continuous deployment, gradually rolling out plugins in ChatGPT to study their real-world use, impact, and safety challenges. This helps the developers to learn more and achieve alignment challenges towards their mission of human–AI interaction. The tool offers a small set of users currently, and larger-scale access is gradually rolling out. Plugin developers can build a plugin for ChatGPT by using the tool's documentation, which then lists the enabled plugins in the prompt shown to the language model as well as provides documentation to instruct the model on how to use each. ChatGPT currently hosts two plugins, a web browser, and code interpreter, and has openly shared code for a knowledge base retrieval plugin. Users and developers can join the ChatGPT plugins waitlist and have access to plugins gradually over time. The initial alpha access will prioritize a small number of developers and ChatGPT Plus users, and gradually expand over time. The aim behind plugins is to unlock a vast range of possible use cases and build a community shaping the future of the human–AI interaction paradigm.
F.A.Q
Browsing by Open AI is an attribute of ChatGPT which enables it to access the most recent information, execute calculations, and utilize third-party services via plugins. One of the plugins hosted by ChatGPT offers a web browsing function, allowing the AI system to interact with and retrieve data from the internet.
The plugins in ChatGPT are designed to increase its range of capabilities, specifically enhancing its access to up-to-date information, its ability to perform computations, and its use of third-party services. This helps in augmenting the utility of the system and allows it to cater to a variety of use cases. The ultimate aim is to unlock a vast array of possibilities and shape the future of human-AI interaction paradigm.
ChatGPT has implemented several safeguards at the core of its plugin platform to maintain safety. It prioritizes safety in its plugins design and particularly in accessing the web, it operates within a safe browsing environment, respects website requirements for bots and maintains strict controls to prevent external internet access from executed code. Developers are also encouraged to integrate only content they are authorized to use to ensure data authorization and privacy.
ChatGPT can access various types of information with its plugins. With the web browsing plugin, it can access and retrieve up-to-date and specific information from the internet. With the code interpreter plugin, it interacts with a Python interpreter. There is also a retrieval plugin that can access personal or organizational information sources, allowing ChatGPT to retrieve document snippets from data sources like files, notes, emails, or public documentation.
The continuous deployment of plugins in ChatGPT means that new plugins and updates to existing plugins are progressively rolled out. This strategy allows the developers to test, observe, and refine plugins in a real-world context, studying their usability, impact, and potential safety challenges. This iterative deployment philosophy helps in ensuring a balance between innovation and system stability.
Users can gain access to ChatGPT plugins by joining the ChatGPT plugins waitlist. Access to plugins is being gradually rolled out over time, with priority initially given to a small group of developers and ChatGPT Plus users.
To build a plugin for ChatGPT, developers who have been invited off the waitlist can follow the tool's documentation. A plugin includes a machine-readable description of its capabilities and how to invoke them, as well as user-facing documentation. Once built, the plugin will appear in the listed enabled plugins for the language model and usage instructions will also be provided.
ChatGPT currently hosts two plugins, a web browser and a code interpreter. Additionally, they have shared the open-sourced code for a knowledge base retrieval plugin, allowing developers to create their own version and augment ChatGPT's capabilities.
The ChatGPT plugins waitlist is a system by which users and developers express their interest in gaining access to ChatGPT plugins. As the roll-out of plugins is a gradual process, those on the waitlist will gain access over time.
ChatGPT aims to use plugins as a pathway to unlock a wide variety of use cases and contribute to the evolution of the human-AI interaction paradigm. The usage of plugins enhances the usability of the system, allowing language models to access recent information, perform computations, and use third-party services, thereby leading to more meaningful and dynamic interactions.
Through its plugin deployment, ChatGPT aims to study the real-world use, impact, safety, and alignment challenges. The deployment provides necessary insights into the system's effectiveness, its potential risks, and the overall user experience. This information assists in the continuous refinement and evolution of the tool, ensuring it stays aligned with their mission of enhancing human-AI interaction.
The initial alpha access of ChatGPT plugins will be given to a small group of developers and ChatGPT Plus users. This approach allows for effective testing and refinement of plugins before they are made available on a larger scale.
The retrieval plugin for ChatGPT can handle a variety of documents. It allows users to obtain relevant document snippets from their data sources such as files, notes, emails, or public documentation. This means the retrieval plugin can access and summarize information from a wide range of textual sources.
The code interpreter plugin in ChatGPT allows the model to interact with a Python interpreter in a sandboxed, firewalled execution environment. It supports the execution of Python code, handling of file uploads and downloads, and provides ephemeral disk space. The plugin evaluations of executed code build on top of each other, supporting complex tasks and helping create new workflows.
While navigating the web, ChatGPT uses a text-based web browser plugin that respects websites' robots.txt files, avoiding crawling of sites that have prohibited it. Also, the browsing activities are isolated from the rest of the infrastructure, and the plugin makes only GET requests, limiting potential security risks. The user-agent token, ChatGPT-User, identifies the browsing activities of ChatGPT and is configured to honor the crawling restrictions of websites.
Third-party plugins in ChatGPT offer a range of features, extending the capabilities of ChatGPT to perform specific tasks or use certain services. These plugins may provide functionality like restaurant reservations, computation and math capabilities, grocery delivery, language tutoring, and integration with various apps. The features of third-party plugins greatly expand the potential use cases of ChatGPT.
Third-party plugins in ChatGPT are documented through a manifest file which includes a machine-readable description of the plugin’s capabilities and how to invoke them. The file also includes the user-facing documentation which offers guidelines on how the plugin can be employed by the user. This detailed documentation aids both the model in understanding how to use the plugin and the user in knowing what to expect from it.
Yes, developers can create their own plugin for ChatGPT. Developers who receive an invitation off the ChatGPT plugin waitlist can follow the guidelines provided in the ChatGPT documentation to build an API with endpoints for the language model to call. They can create an OpenAPI specification to document their API, and a manifest file that includes plugin-specific metadata as well as a link to the OpenAPI spec.
The manifest file in ChatGPT plugins serves as the written documentation of the plugin's capabilities which is understood by the language models. It includes a machine-readable description of the plugin's capabilities and how to invoke them, along with user-facing documentation that explains the functionality of the plugin to the user.
Multiple organizations have contributed plugins to ChatGPT, including Expedia, FiscalNote, Instacart, Kayak, Klarna, Milo, OpenTable, Shopify, Slack, Speak, Wolfram, and Zapier.
Pros and Cons
Pros
- Browsing capabilities
- Safety as core principle
- Continuous deployment
- Gradual roll-out
- Comprehensive documentation for developers
- Potential for diverse plugins
- Community-building potential
- Web browsing integration
- Gradual alpha access
- Expands utility of language models
- Capability to perform actions
- Possibility of open standards
- Improved user engagement
- Information retrieval plugin
- Python interpreter plugin
- Good for mathematical problems
- Data analysis and visualization capabilities
- File format conversion
- Support for third-party services
- Designed for language models
- Real-world use impact analysis
- Cites sources in responses
- Supports GET requests
- Sandboxed
- firewalled execution environment
- Ephemeral disk space
- Persistent python session
- Supports file uploads and downloads
- Internal and external red-teaming
Cons
- Limited plugin support
- Gradual plugin roll-out
- Safety measures restrict functionalities
- Alpha access bias
- Limited scale access
- Plugin development dependent on documentation
- Limited hosted plugins
- Gradual waitlist access
- Exclusive plug-in developers
- Process for new plugins unspecified
Reviews
You must be logged in to submit a review.
No reviews yet. Be the first to review!
Similar Tools
Document collab w/ chat in multiple langs & formats
TBD
Free + from $14.99/mo