Replies: 2 comments 1 reply
-
I was thinking about the combination of HuggingGPT and AutoGPT, I think you are describing the same thing? |
Beta Was this translation helpful? Give feedback.
-
I have chatGPT3.5t in my api key maybe you get then when you subscribe. you also get chatGPT4 which is better at coding and makes way less mistakes. Also supports Open AI if you have money otherwise you could try to help to implement/add vicuna, alpaca 13b, gpt4all usw. (free language modes that got waaaaay better but still not as good as chatGPT open ai models as far as im aware of/guessing. Here lots of L.m. compared: https://chat.lmsys.org/ Also if there was some way for the community to help autoGPT development without messing anything up it would be great. I have many ideas and plans but i cant contribute them. I can also minimally code and have chatGPT4 access. Other people might have no creativity or clue at all and be genius coders others just good and organisation idk what. Some maybe just good mood. I also have no idea how AutoGPT works since im too low drive to go trhought all the code. So someone could do that. Idk I already raised those suggestions in the last days in some posts. 👍 |
Beta Was this translation helpful? Give feedback.
-
I've been exploring this for a few days now (using only GPT-3, as I'm still awaiting GPT-4 access), and these are my findings:
The current issues and pull requests overflow with external features and services, making it challenging to manage a monolithic application. Considering AutoGPT's inherent delegation-based approach, it is counterproductive to consolidate all these features into a single package. Instead, AutoGPT should emphasize a streamlined design that efficiently delegates tasks to plugins or addons when necessary.
The developers could potentially maintain a repository of "official" plugins while also allowing anyone to create their own. AutoGPT could then automatically review the available addons and choose the most appropriate ones based on its needs.
Implementing this approach would enable AutoGPT's capabilities to expand exponentially without overburdening the repository maintainers, who are already swamped with pull requests.
Beta Was this translation helpful? Give feedback.
All reactions