Couldn't agree more lol
While human hallucinations may only have a limited impact, the implications of AI model hallucinations are uncontrollable given their widespread use in daily life.
I totally hate the ontology of AI models. It's a constant for me: allowing these words to be come terms of art invites the mis-application of meaning into ordinary discourse.
Whatever it is they do, doing it less or more than humans isn't the point. They aren't humans, and its not a human behaviour, and trying to sneak it in as "like" is really a sin of commission: it's part of the message to make personhood assumed not tested.
Do people have rich fantasy life? Do they routinely mis-read and mis-hear, and mis-see? Sure! Thats what a mondegreen is. It's what a whole bunch of Insurance accident reports are: you literally didn't see the other car.
We functionally live inside a simulation of the world, created internally by the brain. we all know that. So at one level the CEO is being disingenuous.
I'm thinking about a carrot right now. it has a face. and hands. It's attached to the CEO of Anthropic in several entertaining ways. Does the carrot exist? No: it's a conceptual carrot. But I sure am thinking about it.
Thanks for the insightful and engaging sharing. I believe that the human brain, along with our cognition and behaviour, is truly the most remarkable creation. The distinct boundaries between humans and AI should not be blurred.
"AI models hallucinate less than AI company executives"