Recent API update gives ability to do function calling

[pawel-dubiel]

https://platform.openai.com/docs/guides/gpt/function-calling
So exited this should be ground breaking. That should make all the models much more reliable in AUTO-GPT

GPT-3.5 Turbo

This model has been updated with a new version: gpt-3.5-turbo-0613 which is more steerable with the system message and includes a new capability: function calling. By describing functions in your prompts, the model can intelligently output a JSON object containing arguments to call these functions based on user input — perfect for integrating with other tools or APIs. Learn more in our function calling documentation

Plus enjoy a 25% cost reduction for input tokens on GPT-3.5 Turbo (now $0.0015 per 1K input tokens), effective immediately.

Longer Context

We’re also introducing gpt-3.5-turbo-16k. This model offers four times the context length of the 4k base model and is priced at $0.003 per 1K input tokens and $0.004 per 1K output tokens.

Model Transitioning

You can begin using the new gpt-3.5-turbo-0613 model today. On June 27th, the stable gpt-3.5-turbo will be automatically upgraded to this new version. If you need more time to transition, you can specify gpt-3.5-turbo-0301 to keep using the older version, which will remain available until September 13th as part of our upgrade and deprecation process.

[GoMightyAlgorythmGo]

Can function calling be used to call previously made stuff aka improve the models longterm planing ability like in the voyager paper? Or is it smth else?

[GoMightyAlgorythmGo]

The context window increase is huge, but there seems to be some issue with AutoGPT where its basically not working anymore anyway whenever i tryed it int he last months, i mean it should already work if you plan correctly but does not seem to work loops/errors so cant tell if the planning got better but to be it was a lot better in ~15 april