Well, I’ve made some progress in my AI experiment. As you can see from this GIF that Nova is capable of subtle movement while waiting for something to happen. However that’s not really what people come to see.
So lets talk about the changes from my last post until today.
First off, Nova no longer relies on JSON or PowerShell to handle memory management. In fact, I’ve changed the base language in favor of another that is lighter, faster, and more compatible with other system resources and would allow more flexibility.
Yes, I’m purposefully being vague. However the changes resulted in a faster bootup time and quicker replies. It also allows her to search quickly for a memory instead of queueing the entire file, reading it from start to finish, then forming a reply. All of that bloat to her active memory was causing her to slow down as she learned so this had to be done.
While I was changing languages I kept running into the same issue over and over. I fix her memory, her renderer decides to try and hold her runner hostage. I fix the renderer and now the runner decides it doesn’t want to run. Fix the runner.. you get where I’m going with this. The final solution.. Give the renderer its own window. Instead of fighting it.. I’m going to turn a bug into a feature. Which I’ll be honest isn’t the biggest deal since this also gives more room to Nova to be expressive.
It also gives me more room for a larger chat history and more buttons to add. So I’ll not take this as a loss of sanity and more of a gain in real-estate.
So what else changed? Well for starters I added a button that allows me to read Nova’s mind a bit better. Which is really more for me to understand when things go wrong.
I’ve also added a few tools that help me to understand what happened when her animations fail.. such as her early wave. Nova would point at the chat where she wrote “wave” instead of animating a wave.
The biggest upgrade, is something that is still a work in progress.
I’m breaking the subconscious routine into two, mostly because I need to understand what is failing.
SubCon: Handles all the things she doesn’t need to actively think about.. Like a human doesn’t need to think about their heart beating or breathing in order for those things to happen.
Exec: These are things that the main Conscious needs to know about, but doesn’t necessarily need to run on the heaviest loading part of the brain. It helps with planning and allows multiple related tasks to be handled at once.
Which is the problem.. that.. and it’s not the endgame for this routine either. In fact.. it’s whole purpose is really to stop Nova from having to rely on input in order to act. If I can figure out how to properly code this then the next step is polishing her vision and giving her access to see the screen.