>>769
Hey man, just an occasional lurker here, I pop in every now and then when I have a rare moment of free time. I don't have much in the way of technical skills to add much to the conversation all my degrees are in humanities related subjects, (currently working on a Master's degree in history.)
Still, always been fascinated with robotics, Megaman was my childhood and Chobits was my jam back in Highschool, I've enjoyed watching y'all's progress and found some of the discussions rather fun to read. I particularly liked >>>17125 reading your Robowaifu Ethics and Morality thread Chobitsu.
Anyway, just wanted to briefly break my silence to say that I am relieved that things are back up, I had been worried there for a moment, and that I am cheering on y'all's work. God bless.
Hmm, actually, while I am here, I got a question but I am not sure if these topic has been discussed elsewhere or not, or if it might give away my lack of knowledge on the subject of AI/robotics.
1. What is this board's opinion on AI agents? I've used AI a lot such as feeding research material into NotebookLM and Perplexity then asking the PDF questions as a time saver, and I recently saw where some people have managed to automate the process doing research, having a team of AI scrape the internet, analyze the data, then present the data complete with citations. On the other hand, there was a study done where bees have been observed briefly pausing work to plays with balls placed in an enclosure. Here is a creature with a brain the size of a grain of sand displaying playfulness. Minimal compute power, maximum results. My thoughts are, maybe with an eye towards biomimicry, save on compute power needed for your robowaifu by composing her mind out of many low compute specialized ai that break the task down of analysing sensor data, pulling up relevant memories, and choosing a course of action into bite sized pieces. For example, an AI designed to manage the touch sensors detects something past the limits of material might bypass the decision making process by directly sending an order to quickly move the pained limb away from the direction of the pain source. That's a reflex reaction with barely any compute used. Decant situational context and relevant memories down enough and it might cost less tokens for the conscious portion of the AI to make decisions. Anyhow, this is not my area of specialty so it won't hurt my feelings if it turns out to be something that was already discussed or something not feasible.
Anyhow, I'm just glad things are starting to get up and running again.