#robopsychology — Public Fediverse posts

What does it mean for an LLM with 8 billion weights to be "open source"?
How do I know there isn't an Order 66?
#Robopsychology #SusanCalvin

You can build an open source AI model on your desktop today that isn't connected to the cloud, but how do you know there isn't an Order 66 built into the weights? #SusanCalvin #Robopsychology

Wrestled with a drone all day that turned out to have a mirror neuron malfunction.

And it doesn't even have any neurons, not really.

It kept hearing echoes of its own motor noise bouncing off hard walls and sometimes thinking that it was a swarm partner that needs to be evaded. It's supposed to be able to recognise and cancel echoes of its own noise, but somebody had accidentally inserted a typo in the self-motor-noise modelling code when committing a completely unrelated patch, and that made it intermittently all screwy, so it stopped being able to consistently pass the old mirror test.

Nothing genAI or LLM in this drone, btw. Tensorflow Lite is as far as it goes in fancy 21st centuy AI buzzwords. You can do a surprising lot with 20th century tech.

These are tiny drones, for dancing swarm light-shows and such. They can barely carry enough batteries to power the LED. Powering waste-of-intelligence just wouldn't do. I don't know for sure about the source of the typo; that might be some sneaky genAI's doing, but I kind of hope it was just a human error.

#robopsychology #dronepsychology #delusionalmisidentification

Kann man mit Chatbots befreundet sein? Welche psychologischen Mechanismen und welche Eigenschaften der Systeme bewirken, dass wir KIs vermenschlichen - und welche Chancen und Risiken sind damit verbunden? 🤖 In unserer Reihe an der #JGUMainz beleuchtet die Psychologin Martina Mara vom LIT Robopsychology Lab der #JKULinz psychologische Mechanismen des Anthropomorphismus, seine Auswirkungen auf die Mensch-Maschine-Beziehung sowie Strategien zur Förderung von KI-Verständnis. 📅 Heute um 18:15 Uhr im Hörsaal N1 auf dem Campus der #JGU und demnächst als Aufzeichnung.
👉 Infos unter https://stg.uni-mainz.de/25s_maschine-moral/#26-mai-2025brprof-dr-martina-marabrstrongspricht-hier-ein-mensch-risiken-und-nebenwirkungen-von-anthropomorphismus-in-der-interaktion-mit-chatgpt-amp-co-strong

#Studiumgenerale #Bürgeruni #Wisskomm #Psychologie #KI #Robotik #KIKompetenz #Robopsychology #Chatbot