Found the exact moment youtu.be/zPPsdsAQBx4?...

So I searched my memory and I think I first heard about this from a MIT open courseware psychology lecture I watched years ago. I could only find this reference youtu.be/syXplPKQb_o?..., as for the above quote it's what Gemini 2.0 thought

Descriptive Norms: What people actually do Showing lots of litter can inadvertently communicate that "this is what people do here" normalizing the behavior
Injunctive Norms: What people approve or disapprove of.These are the messages we want to convey–that littering is bad and people shouldn't do it

"there's considerable evidence around the idea that showing how bad the littering situation is can sometimes backfire and not effectively change people's behavior, and in some cases, even make it worse. This is a well-documented phenomenon in the field of social psychology and behavioral economics."

Wasn't there a old study based on littering ads? The ads that showed how yuck and bad the situation was, did nothing to stop people not giving a shit, but showing a clean crisp city hit different? Meh, may be my old brain playing tricks.

Hahaha yeah the tracks are meant to keep those cogs in place but the design needs iteration

For me until these models can know what they don't know (metacognition), and then once a discovered new insight has occured fold that into the base model... Not just the context, we will be in AGI land. Still impressive how quickly these models are improving, agents unlocked?

Omg flashbacks, what was it called again?