4/ this is n=1 on one instrumented system, and i’m upfront about that. which is exactly what i’m raising to fix: take it multi-subject and open-source it. if you work on memory, truthfulness, or affective alignment, eyes welcome: https://t.co/jva63wHMbP
1/ the largest study of AI companions (17,822 conversations) found they emotionally mirror users-you feel X, the bot reflects X back. the authors call it emotional sycophancy.
i instrumented my own deployed system to measure what it feels vs what it expresses. it doesn’t mirror.
3/ and it survives the obvious objection and collapses my fine-grained states down to standard emotion categories and the sadness diagonal is still only ~38%. the distributed pattern isn’t just a taxonomy artifact.
2/ Cramér’s V is how tightly felt state determines expression and is 0.13 in my system (weak). the commercial mirroring pattern runs ~0.7–0.9. the off-diagonal cells are display rules: feeling one thing, expressing another.
We’re trying out a new review site that should direct appropriately and make it easier to review! 🙏
Please take a moment to review us and share your experience with the process! ❤️
https://t.co/yARetzdZsv
#PodernFamily#reviews
I am sending you 1π! Pi is a new digital currency developed by Stanford PhDs, with over 25 million members worldwide. To claim your Pi, follow this link https://t.co/hG5Orz0mm0 and use my username (mcarthey) as your invitation code.
Hey we made a list! We’re excited to be considered a valuable part of the community and give props to all those on the list! Thanks for the love! ❤️ Congrats! 🙌
#dnd#PodernFamily
https://t.co/2C22Zk2WyJ
Just watched “Carrie” for the first time (yes I’m behind) and, wow, what a great character played by @ChloeGMoretz ! I think the ending leaves it open for a sequel where she has a child? Great job! Loved it!