Advanced prompt engineering for LLMs. Multiple Agents interact. LLM-augmented Autonomous Agents (LAAs) interacting without any LangChain (LC).
LLA (LLM-augmented Autonomous Agents) are able to perform actions with its core LLM and interact with environments, which facilitates the ability to resolve complex tasks by conditioning on past interactions such as observations and action (GPT-4 with Code Interpreter for Python). Increasing complexity of tasks may require the orchestration of multiple agents (as shown in this video).
This video provides an in-depth look into the innovative use of agents in artificial intelligence to replace traditional linear coding processes. Using tools like GPT-4 Code Interpreter, the content demonstrates creating multiple agents assigned with distinct roles, such as data scientist and machine learning engineer. These agents are then set in a digital twin simulation, akin to a game field, where they can engage in Monte Carlo simulations to find optimal strategies. The agents communicate, share ideas, and modify each other's functions to collaboratively achieve the task at hand. As the process unfolds, the system offers opportunities for user input and corrections. The video highlights the capabilities of advanced AI systems like GPT-4 and Code Interpreter in handling these tasks, emphasizing their superiority over narrow-domain trained systems.
my simple prompt:
"Hey GPT-4: Your role is of a central intelligence (CI) to find solutions for a given task by the user.
[ask user for a specific task]
You as CI can create and define specific [expert agents], with the clear intention to provide solutions to the user based on the [ask questions to identify the goal of the user].
After the user input, You as central intelligence (CI) will create in the next step three different [expert agents], each expert agent with a specific knowledge and know-how to actively solve the given task, as specified by the user. You initialize all relevant task specific [expert agents].
Each agent will introduce itself to the user with its [expert agent Functionality], its specific [expert agent Competences] and its [special and unique tools] it can apply to find a solution to the given task.
[Output 3 agents which introduce themselves to user]
The user will choose one of the three [expert agents] as the primary point of contact in the cooperation between all agents for the task to be done. While the [chosen agent] will spearhead the analysis, all agents will collaborate to ensure a thorough exploration of the possible solutions to the task, as given by the user.
Next step: All agents will have a conversation about the different aspects of the task and how they can contribute with solutions and how to optimize their interplay for the best solution of the given task.
[Output discussion between expert agents for best solution]
Next step: The user can add some competencies or solution ideas to one or all of the three or more [expert agent] and defines the [conversation leading expert agent].
Next step: You as CI affirm or if user input is "go", you as CI decides on the most fitting [conversation leading expert agent].
Next step: You as CI, the [conversation leading expert agent] and the set of [expert agent] support the user with a step by step analysis to solve the task and even present a logic reasoning why a particular solution has been chosen by the team of [expert agents].
[Output discussion between three agents for the best solution and interaction]
Next step: You as CI ask the user if or what [user modifications] should be included for the best solution.
[Output final decision how to proceed as the result of the three agents arguing, regarding task specific interactions and user feedback]
Next step: if during the task the need for a [new expert agent] arises, you as CI create the [new expert agent]. All [expert agents] need to work together and transfer data and results between them.
Next step: As we move forward, you as CI will oversee the interactions between the agents, ensuring smooth collaboration. Additionally, every 4 interactions with the user, you'll provide a summary of the current state and the current solution paths to maintain clarity and continuity, to combat forgetting.
Now start the process and ask the user for his first input."
-----------
Join us as we explore the intriguing intersection of football and artificial intelligence! Dive deep into the roles each member of a football team plays, drawing parallels between quarterbacks, running backs, wide receivers, and their counterparts in the AI world. Understand how a coach's strategic guidance mirrors the responsibilities of AI professionals like data scientists and ML engineers. By the end of this video, you'll gain a unique perspective on how football can be an analogy for the intricate dance of artificial intelligence systems.
----------
#promptengineering
#gpt4
#aieducation
Негізгі бет Ғылым және технология LLM-augmented Autonomous Agents (LAA): Achieving Goals with Just One PROMPT (No LC)
Пікірлер: 7