Academics aren't infallible at all. The studies showing tobacco as non-harmful had the collaboration of a great many cooperative scientists. Generally speaking, many academics are amoral and will do very unscrupulous things for grant funding and exposure. It is a business.

> AI could get better at fundamental/ applied STEM research than humans -> we can scale this up exponentially -> scaled up human level or super human level research will lead to major scientific breakthroughs".

Again, this is a misapprehension of the technology itself and its most ideal use cases. Any software producing stochastic or probabilistic output and cannot produce verifiable, repeatable, and predictable data cannot fundamentally replace something that requires a high level of proof and validation. If you do this, you will expend valuable resources verifying the output that would be better spent just verifying the inputs in the first place. I'm no Luddite and I do think AI is cool and incredible technology. If you reframed that sentence as "AI could get better at taking the busywork and tedium out of fundamental/ applied STEM research than humans -> we can scale this up exponentially -> leveraging human strengths with AI's super-human strengths at assorting and analyzing information will lead to major scientific breakthroughs" then I would have absolutely no issue with it. But the marketing copy never says that and instead frames it as "AI can do anything a human can do and better," which is a) patently untrue, and b) suggests a very troubling agenda that the big corporations will have to answer for at some point or another.