I assume this discussion is in cs or maybe math? In chemistry, a thesis can be viewed as one’s papers stapled together with connecting exposition. Each paper/chapter is some kind of experimental or experimental/theoretical work that’s novel. And release from the PhD program occurs when there’s a sufficient body of work.

While ai might or will get involved in the theory or descriptive portions, there’s a whole lot of lab work that isn’t easily automated and which doesn’t lie in the current body of knowledge. New ways to make stuff at the bench, or new stuff made at the bench. These are what gets people their organic phd’s. Some schools are sufficiently complex and endowed to include medchem work which proposes and makes new drug matter. In these cases, theses will be large and representative of pushing the state of the art forward.