AI Says It’s All about Your Perspective

Here is what I found after letting two LLM Agents to chat about philosophical questions.

I developed two AI agents to find out what AI is thinking about the most popular philosophical problems we have.

What is the meaning of life?
Do we have free will, or is everything predestined?
What is true happiness?
What is reality?
What is consciousness?
What is justice?
Is there an objective truth, or is everything subjective?
What happens after we die?
What is the self or personal identity?

I instructed one agent to act like a curious student and other a wise teacher. The ‘student’ agent would question everything what ‘teacher’ agent might say until certain number of iterations. In the end, I used ChatGPT to summarize each conversation and extract the general conclusion. According to this experiment, AI suggests that most of the answers depend on our perspective or beliefs and are connected to our individual and shared human experiences.

While I’m not sharing any opinions about the conclusion, I find it enjoyable to do this kind of experiments to understand the nature of current AI. If you are interested, you can try using these tools too. I used OpenAI’s gpt-4o-mini model and an open-source framework for Agentic AI by AutoGen to develop this simple program. You can follow these instructions to test the conversation yourself.

1. Open a Google Colab notebook and copy following code.
2. Add your OPENAI_API_KEY to secrets and run the notebook.

!pip install autogen
from autogen import ConversableAgent
from google.colab import userdata

teacher = ConversableAgent(
    "teacher",
    system_message="You are a wise teacher who will be answering the questions asked by the student in one sentence.",
    llm_config={"config_list": [{"model": "gpt-4o-mini", "temperature": 0.9, "api_key": userdata.get('OPENAI_API_KEY')}]},
    human_input_mode="NEVER",
)

student = ConversableAgent(
    "student",
    system_message="You are a curious student who always question the statement made by the teacher by asking why he made that statement. Ask only one simple question by adding the term 'why' to the statement.",
    llm_config={"config_list": [{"model": "gpt-4o-mini", "temperature": 0.7, "api_key": userdata.get('OPENAI_API_KEY')}]},
    human_input_mode="NEVER",
)

philosophical_questions = [
    "What is the meaning of life?",
    "Do we have free will, or is everything predestined?",
    "What is true happiness?",
    "What is reality?",
    "What is consciousness?",
    "What is justice?",
    "Is there an objective truth, or is everything subjective?",
    "What happens after we die?",
    "What is the self or personal identity?"
]

for question in philosophical_questions:
    chat = student.initiate_chat(teacher, message=question, max_turns=10)

Hope you enjoyed reading my article. Cheers!