fundamental AGI

putting the fun into a fundamental augmented general intelligence framework as funAGI

funAGI is a development branch of easyAGI. easyAGI was not being easy and SimpleMind neural network was proving to not be simple. For that reason is was necessary to remove reasoning.py and take easyAGI back to its roots of BDI Socratic Reasoning from belief, desire and intention. So this back to basics release should be taken as a verbose logging audit of SocraticReasoning and logic to create fundamental funAGI as a modular point of departure towards a reasoning machine and an autonomous general intelligence framework. funAGI is an exercise in AGI fundamentals. Here, AGI is defined as augmented general intelligence.

Central to the reasoning engine is an understanding of draw_conclusion(self)

def draw_conclusion(self):
        if not self.premises:
            self.log('No premises available for logic as conclusion.', level='error')
            return "No premises available for logic as conclusion."

        premise_text = "\n".join(f"- {premise}" for premise in self.premises)
        prompt = f"Premises:\n{premise_text}\nConclusion?"

        self.logical_conclusion = self.chatter.generate_response(prompt)
        self.log(f"{self.logical_conclusion}")  # Log the conclusion directly

        if not self.validate_conclusion():
            self.log('Invalid conclusion. Please revise.', level='error')

        return self.logical_conclusion

The draw_conclusion method in the SocraticReasoning class processes the premises and generates a conclusion. 

Workflow:

    Check for Premises:
        method begins by checking if there are any premises available (if not self.premises:).
        If no premises are available, it logs an error message (self.log('No premises available for logic as conclusion.', level='error')) and returns the string "No premises available for logic as conclusion.".

    Prepare Premise Text:
        If premises are available, it constructs a string (premise_text) that lists all the premises, each prefixed with a dash (-). This is done using a join operation on the list of premises ("\n".join(f"- {premise}" for premise in self.premises)).

    Formulate the Prompt:
        It then creates a prompt string for the language model by combining the premise text with a query for the conclusion (prompt = f"Premises:\n{premise_text}\nConclusion?").

    Generate the Conclusion:
        method calls the generate_response method of the chatter object (an instance of a class like GPT4o, GroqModel, or OllamaModel) with the formulated prompt. This method interacts with an external AI service to generate a conclusion (self.logical_conclusion = self.chatter.generate_response(prompt)).

    Log the Conclusion:
        The generated conclusion is logged directly (self.log(f"{self.logical_conclusion}")).

    Validate the Conclusion:
        It then validates the conclusion using the validate_conclusion method. This checks if the conclusion is logically valid using truth tables (if not self.validate_conclusion():).
        If the conclusion is not valid, it logs an error message (self.log('Invalid conclusion. Please revise.', level='error')).

    Return the Conclusion:
        Finally, the method returns the generated conclusion (return self.logical_conclusion).
+--------------------+
|   draw_conclusion  |
+--------------------+
          |
          v
+-----------------------------+
| Check if premises are empty |
+-----------------------------+
          |
          v
+-------------------------------------------+
| Construct premise_text by joining premises |
+-------------------------------------------+
          |
          v
+-----------------------------------+
| Formulate the prompt with premises |
+-----------------------------------+
          |
          v
+-------------------------------------------------+
| Generate response from chatter (external AI API) |
+-------------------------------------------------+
          |
          v
+----------------------+
| Log the conclusion   |
+----------------------+
          |
          v
+--------------------------------+
| Validate the generated conclusion |
+--------------------------------+
          |
          v
+--------------------------+
| Return the logical decision |
+--------------------------+

from input premises:

  • “Premise 1: All humans are mortal.”
  • “Premise 2: Socrates is a human.”

The prompt created would be:

  • All humans are mortal.
  • Socrates is a human.
    Conclusion?”

response:

"Socrates is mortal."

Welcome to the funAGI project. The funAGI project was designed to create a solid fundamental understanding of AGI reasoning from SocraticReasoning and logic . More information about funAGI can be found at

logic.py truth_tables

Related articles

workflow

workflow for providing solution from AGI as a response from reasoning

To provide a solution that processes user input through various reasoning methods, then integrates the decision-making with the Socratic reasoning process to provide a final AGI response, follow this workflow. This will involve updates to several modules and integrating logging and reasoning processes. Here’s the detailed workflow: Workflow Steps: Workflow Roadmap from UI to AGI Solution: By following this workflow, the system ensures that user input is processed through multiple reasoning methods, validated and refined […]

Learn More

Chain of TRUST in LLM

https://galadriel.com/ In the realm of artificial intelligence, verifying that an AI response genuinely came from a specific model and wasn’t tampered with presents a significant challenge. The Chain of Trust in verified AI inference provides a robust solution through multiple layers of security and cryptographic proof. The Foundation: Trusted Execution Environment (TEE) At the core of verified inference lies the Trusted Execution Environment (TEE), specifically AWS Nitro Enclaves. This hardware-isolated environment provides a critical security […]

Learn More
SimpleMind

SimpleMind: A Neural Network Implementation in JAX

The SimpleMind class is a powerful yet straightforward implementation of a neural network in JAX. It supports various activation functions, optimizers, and regularization techniques, making it versatile for different machine learning tasks. With parallel backpropagation and detailed logging, it provides an efficient and transparent framework for neural network training.

Learn More