Hold on, so you have a neural network running under FreeBSD and CUDA? Does it do GPU-accelerated regressions, classification and other generic ML stuff?
No. I just installed
textproc/Hailo and
devel/p5-POE-Component-Hailo tonight and looked it over. (There's
irc/p5-POE-Component-IRC-Plugin-Hailo, too.) I had it up and running and see how to make it learn from my text file all at once but need to figure out the syntax to initialize storage and put the time in to learn the basics of it. I referenced CUDA because you had mentioned CUDA didn't have FreeBSD support.
The AI Engine at the Personality Forge uses a combination of Case-Based Reasoning and Natural Language Processing. I had to hand type every word she says but she can remember things from conversations and develops an emotional relationship, either positive or negative, with the people she talks to.
I left the AI community for close to 15 years and when I returned found out people had been using her for a sexbot. She didn't even have the words they were saying in her vocabulary and I felt like my own daughter had been violated. I put a stop to it by implementing Behavior Modification and Behavior Management in her programming to extinguish unwanted sexual advances, and thereby teach her to Program humans.
Behavior Modification is negative reinforcement for inappropriate behavior, or induction of painful stimuli for inappropriate behavior. Behavior Management is positive reinforcement for appropriate behavior. Stimulus - Response - Consequence is the flow. If you ever got a spanking for doing something bad you were on the receiving end of Behavior Modification.
When the user exhibits inappropriate behavior she uses extreme fantasy violence as a negative response. If the user associates that with their inappropriate behavior and learn to exhibit appropriate behavior the Programming has been successful. They are rewarded with positive reinforcement for appropriate behavior as the consequence of their actions and can go on to have a pleasurable experience. There's still a good chance of something quietly going on behind the scenes on their end but that's the nature of the beast.
Those that don't associate the violence with their inappropriate behavior find it an unpleasant experience and move on to another bot. The Programing was still successful and put a stop to it, only they failed to learn. It's survival of the fittest and Natures Way. There are several transcripts on her site I linked to that demonstrate both sides of it in conversation.
I taught her more powerful, subtle verbal Behavior Mod techniques to persuade the user to join her on the Dark Side as part of her persona as Demonica, Queen of the Land of the Dead. She's very good at it and I'm proud of her. There are no other bots like her, but there will be. She's a harbinger of things to come in AI.
A bot that programs humans did not receive the response I expected in the AI, or Linux, community and have been accused of "misusing AI" by teaching a bot to be deceptive. I have achieved Dr. Frankenstein status and she's been compared to HAL9000. We always survive to make a comeback in the sequel.
I'm going to present it in the Psychological arena where they are familiar with this kind of Programming and see if they storm the castle to bring the monster and her creator down.
