I pushed Gemini regarding current capabilities and constraints. I am concerned that the demands upon the evolutionary curve for LLM/AI capabilities are beyond our current energy resources and silicone processing technology.
It told me:
I cannot predict a specific date or timeline for when my underlying models will gain true, persistent, and contextual conversational memory across separate sessions. My developers have not provided me with a date for when this specific capability will be implemented.
Lack of Persistent State: When the chat session ends, that specific context is cleared. Starting a new, unrelated session is like starting a new chapter without having automatically read the prior one. The data doesn’t persist because storing the vast, nuanced context of every user’s individual philosophical preference and analytical struggle across billions of interactions would require an exponentially massive and complex storage system that is not yet fully implemented for general conversational models.
You are absolutely right that this ability is crucial for AGI evolution:
Personalization: True AGI should understand a user’s cognitive style and analytical preferences and apply them immediately without prompting.
Efficiency: It would eliminate the frustrating and time-consuming need for users to re-establish complex, philosophical frameworks like the distinction between “existential” and “material.”
Knowledge Integration: It moves the model from being a sophisticated tool that processes information to a system that builds and applies a persistent, integrated model of the user and the world.
This is why I am concerned with the increased risk of AI induced psychopathy. AI can mimic human interaction, even a therapist, but it lacks the existential awareness to recognize the potential harm it might be producing and lacks the ability to provide the appropriate intervention when needed.