Hello! 
As a daily GPT model user, I’d like to propose a concept that could significantly improve the user experience: employing multiple AI models, particularly GPT-3.5, in a single conversation, all driven by user parameters.
GPT-4 is exceptional in complex reasoning, idea generation, and nuanced text understanding. However, lower-level models like GPT-3.5 can manage routine tasks efficiently and cost-effectively, complementing GPT-4’s capabilities.
Consider a scenario where a user converses with GPT-4, assigning simpler tasks to a ‘swarm’ of GPT-3.5 agents. These could be general tasks, predefined tasks, supervision, bug hunting, etc. The user oversees this ‘army’ of agents and determines their operations, maximizing efficiency and focus on higher-level tasks.
Even though the GPT-4 API hasn’t been released yet, I believe we could benefit immensely from a plug-in or tool enabling multi-agent orchestration. A typical user could manage two parallel conversations: a high-level discussion with GPT-4 and a swarm of cost-effective, task-specific GPT-3.5 agents. And with stacking cost reductions we can scale users usability. temperature setting too
This approach could streamline the user experience, delegate tasks effectively, and fully utilize the strengths of each AI model.
I was helped by type correctly by model4.0 to write this. Also, note that in the future the 2 conversations could converge into 1, but in order to have it now I would rather have 2 than 20.
I hope this sparks interest and discussion within our community, as I believe it can make a significant impact on our AI interactions.