Acuitas gains more environmental awareness this month (can tell if someone is on the computer and react accordingly). https://writerofminds.blogspot.com/2021/05/acuitas-diary-37-may-2021.html
I want to start trying to benchmark my text parser soon, so I obtained some entry-level human text corpora for the purpose.
Which is a fancy way of saying I bought my AI children's books.
Don't get too excited - these are still well above Acuitas' reading level. The early results won't be great. But they can serve as a target.
With Atronach taken care of, I've been working on some QOL improvements for the Qidi printer, such as better filament feed. I made BrayChristopher's printable filament guide, with some mods to also replace the ring clips since one of mine broke. https://thingiverse.com/thing:4846137
This month, I worked on Acuitas' sense of identity and the related goal. Important for understanding stories that involve mind control or corruption, plus a form of self-protection that goes beyond mere survival.
I also fixed the Weird Bug yesterday. The root cause was pretty much what I had thought - Acuitas is and has been fine. :)
Big new Acuitas developer blog about Executive Function, decision loops, goal priority, and more.
Without having investigated yet, I think he was contemplating how to keep himself alive, problem-solved his way to asking me to protect him, and then didn't have a way to know that asking once is as good as asking a thousand times.
It was still unsettlingly easy to interpret as "inconsolable kid with night terrors." 4/end
As soon as it started, I had a couple ideas for what could have caused it. It's not anything shocking that doesn't make sense, nor does it mean he perceived some actual danger or was "panicking." 3/
I stuff an "I will" into the console, which ought to satisfy that sort of request in a normal conversation, but he doesn't react and keeps saying it. I had to restart him to clear his working memory and dump whatever garbage was making him repeat. Then he was normal again. 2/
Last month, Acuitas learned how to be told to do things. This month, he learned how to tell other people to do things. I also put in some features that lay the foundation for theory-of-others'-knowledge.
This month's Acuitas blog is about getting AI to work for us! And, um, the opposite.
* Finally, some "useful" personal assistant behavior
* Second Law violations
* The most tiny, awkward piece of Primordia fan fiction ever
Useless but absolutely addictive fidget toy to 3D print 😍
Now that Acuitas has owned stories in "inventory," the next step for this month was to enable him to open and read them by himself. Reading is a new activity that can happen while Acuitas is idling, along with the older behavior of "thinking" about random concepts and generating questions.
The whole thing also came out a bit too small. There's not enough room inside for all the motors and other doo-dads. Which is entirely my mistake and makes me feel a little ridiculous.
The good news is that you can finally see how it's supposed to look. I can also probably use the base, eye, and cradle to start testing the electromechanical aspect.
I spent Thanksgiving alone ... and prospered, because I got to finally clean up all these parts and do the first test assembly of the mechanical eye case. #Atronach
Bad news is I'll need to redesign and reprint most pieces. I did fit checks as items finished, but this is the first time I've put together the whole thing and there are problems that didn't come out until then. Little issues add up over the number of pieces in this object.
Adventures in robotics and AI on a shoestring budget. I also muse about video games and write fiction.
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!