I fully understand the discourse; my point is that it's not (Nagel, Chalmers, Searle†, and other substance dualists would agree). We need a breakthrough akin to what analysis was to algebra in neuroscience before we even understand what our brain is doing. Whatever it's doing, it's definitely not what OpenAI implemented.
† Yes, I know he says he isn't, but he kind of is[1].
[1] http://www.edwardfeser.com/unpublishedpapers/searle.html
Defining intelligence as "the thing brains do" is a mistake. Planes fly, but not like birds. Biology is a great tool to understand some relationships between form and function, but mimicking biology isn't the end-goal.
This is a pretty bold claim. Brains are our only vantage points, and we also define life as "what's going on in a petri dish." But in any case, your point is self-defeating: what GPT-3 and DALL-E are doing is literally creating models about how we (humans) perceive the world.
Not really. They're creating models of how pixel-space covaries with word sequences over human-meaningful samples. These models are modelling the space of human meaning, not human perception.
Cauchy's work certainly fundamentally changed our understanding of calculus (in that Newton and Liebnitz's earlier hand-wavy proofs were validated with a rigorous definition of a limit), but I'm not aware of any large impact it had on Galois, who was a contemporary of Cauchy.
(Emphasis mine.) Philosophical nit: this by definition would not be AGI. AGI would be the ability of the entity to figure out its own subjective qualia-inducing way of interpreting the world. Thomas Nagel already explained this in the 70s (in his famous essay What Is It Like to Be a Bat?). By enforcing our own worldview (technically speaking, model), we're just making some artificial (and in all likelihood inferior) version of a human.