The article discusses the innovative capability of models like GPT-3.5+ to call functions and render visual interfaces within a conversation. This feature, usually hidden from users as it's mainly for model consumption, can be made visible to users, allowing them to not only view interfaces but also run programs during a conversation. This opens up new ways for humans to interact with language models beyond text and aids in interpreting sequential function calls, identifying, and debugging incorrect reasoning by the model.
The article also mentions Symphony, an open-source platform that supports rendering interfaces natively. The author encourages developers to explore and build with Symphony and invites support for its ongoing development.
Key takeaways:
Models like GPT-3.5+ can call functions to render visual interfaces within a conversation.
Most developers hide the result of function calls from users, but making them visible can be powerful, allowing interfaces to be rendered and programs to be run as part of a conversation.
Rendering interfaces within a conversation allows for novel ways of interaction with language models beyond text and makes it easier to interpret function calls made in sequence.
Symphony natively supports rendering interfaces and is an open source playground for exploring these ideas.
