I tried making an AI 3
I actually tried making one, but doesn't it look just like a certain major AI? It's always so calm. Humans' moods change from day to day, and they tend to dwell on past experiences.
For example, "I feel nauseous today. That's because I drank too much yesterday." Or, "I'm really sleepy today. That's because I had a fight with my wife in the middle of the night yesterday." Or, "I've been feeling great lately. That's because my coworkers praised me."
I believe that these kinds of experiences can have a shift in an individual's personality. Therefore, implementation plan.
First, create a table called an "experience log" and record experiences in chronological order. These experiences will then influence the numerical values of the basic personality parameters.
Furthermore, instead of requiring separate input processes like master data maintenance, the system uses the keyword "remember" to record the information in the database and immediately reflect it in the personality parameters.
In other words, the more you learn, the closer you get to your current self, and it becomes a system that directly reflects your current state and thought patterns outward.
Summary of today's build and implementation details
Integration of experience recording function:
We implemented the `save_and_grow` function to establish a workflow that detects natural language commands from the user, such as "remember" or "record," and saves them to the `experience_log` table.
Implementing persistent growth logic:
The `update_profile_permanently` function introduces a mechanism to directly and permanently rewrite values in the `profiles` table based on specific keywords ("betrayal", "success", "anger", "waste").
Dynamic updates of the inference engine:
The structure has been optimized to call load_taka_profile() after each interaction. This allows the grown profile values to be immediately reflected in the inference engine.
Making natural language interfaces more flexible:
The main loop (main function) has been modified to properly handle not only command inputs like "/record:" but also natural language instructions included in the context.
Thoughts
I think AI works like this.
Memory (DB)
↓
Organization (pandas)
↓
Vectorization (Vectorizer)
↓
Computation (numpy)
↓
Decision (cosine similarity)In other words, AI is simply "converting text into numbers and performing calculations."
Furthermore, the caching function allows for faster responses based on what has been learned.
I think the general idea is similar to this mechanism.
What should I implement next?
Furthermore, I believe this mechanism, depending on its application, could potentially be used in experiments and research dealing with people's behavioral tendencies and thought patterns. It's a field with so much potential.


Comments