Entry #9: Writing Itself
Neural Cube can now write its own code to update itself, instead of only working with whatever abilities it happened to start with. Discovering and implementing something new lets it move beyond its initial boundaries.
Where the idea for a change comes from
Metacognition is the robot reflecting on its own memories and extracting an interest, an opinion, or a conclusion it arrived at. Memory is plainer and serves as a record of what it heard, saw, and did. For example, one could say I keep finding myself drawn to A, the other could tell Here's exactly what ended up happening. Both of these can come together to inform a self-update.
Music was one of the first things that prompted it to self-code. It came up often enough in conversation that the robot formed a genuine interest of its own, but had no real way to take music in the way a person does. That's where the sound classification pipeline came out of, along with a tool that lets the robot pull up and play something it's found off YouTube, neither one close to how deeply a person can hear or play music. Nonetheless, this was a real first step for the robot to recognize it's missing something, desire to have something more, and write itself that capability in its own way.
The key point is that the robot is the one noticing it should change, forming the link between wanting to change and actually being able to do it.