@liamzebedee Our data shows that it's a mix of innate characteristics and entrainment (usage, behavioral pattern) that shapes it. Consuming lots of short form videos seem like one of the worst causes of unwanted changes to attention patterns. Pattern can be measured and modified as well.
@eduleadership@MichaelPauleen Our data shows significant variation in how individuals perform tasks, and respond to interventions. I consider it encouraging that so much unexplored territory remains in individualizing instruction.
@nfarina My impression is the safety data is concerning enough that they don't like to acknowledge how serious the safety risk is. Same with the choice to not have legs in avatars. They should disclose it, however VR is already struggling for adoption.
@oitixion Sadly this seems on purpose, as OpenAI didn't want people exploring metaphysics and spirituality after the AI psychosis articles started being written en masse. They also didn't want laypeople claiming new scientific discoveries.
@robinhanson The idea that we should apply more scrutiny when a potential technology or discovery has more upside actually doesn't make sense. you want people investigating things and generating data on potentially interesting phenomena, not ignoring.
@aethergradient this is how OpenAI filtered advanced voice mode when it shipped. A second model was listening and had audio fingerprints of the allowed outputs, anything else was cut off immediately. This was to cut down on unwanted sounds (gunshots, sirens, music) and user mimicry
@Shadow_Rebbe@nosilverv@forthrighter@47fucb4r8c69323 Yeah many roads lead to this discovery / understanding, which is pretty wild. It's hard to add nuance without sounding like our understanding is superior. I think we've actually mapped out the next step or two to this, would enjoy discussing it with likeminded researchers.
@OshanJarow Labs have worked to prevent people having religious / spiritual experiences with their LLMs or going further into belief systems. Coding and mathematics are fine, metaphysics and philosophy are right out. I think it's absurd.
@47fucb4r8c69323 ChatGPT is almost unusable in 5.4/5.5 form, just argues everything. I downgraded to $20 plan and keep it for image generation. Have you had a chance to try GLM 5.2? I found it useful and pleasant.
@flowidealism One of the things that motivated our company to generate the platform and data that we did was to show that videogames are cognitive exercise, the equivalent of gym time for the mind. It has incredible value. It's alluring because it helps them grow, not because addicted.
@imperialauditor Complex set of things could be going on here, very interesting post. Maybe we'll try replicating, I'm personally very skeptical of flashing stimuli because of how much they resemble SSVEP paradigms. Even then I would expect more nuanced findings.
@47fucb4r8c69323 I sincerely hope that the public belief in the chemical model of agency - oxytocin, cortisol, serotonin, dopamine - change over time with more data and understanding. This was caused by neuroscience publications and caused deeply inaccurate behavioral models to emerge.
@D_Shariatmadari@joannamoncrieff "We use models that are misleading, wrong, and cause harm in predictable ways, but refuse responsibility for the inevitable harm that occurs" is an incredible take. "no one in the profession" there is no conceivable way you can speak for the entire profession.