To quantify it, you'd need measurable changes. For example, if you showed that after widespread LLM adoption, standardized test scores dropped, people's vocabulary shrank significantly, or critical thinking abilities (measured through controlled tests) degraded, you'd have concrete evidence of increased "dumbness."
But here's the thing: tools, even the simplest ones, like college research papers, always have value depending on context. A student rewriting existing knowledge into clearer language has utility because they improve comprehension or provide easier access. It's still useful work.
Yes, by default, many LLM outputs sound similar because they're trained to optimize broad consensus of human writing. But it's trivially easy to give an LLM a distinct personality or style. You can have it write like Hemingway or Hunter S. Thompson. You can make it sound academic, folksy, sarcastic, or anything else you like. These traits demonstrably alter output style, information handling, and even the kind of logic or emotional nuance applied.
Thus, the argument that all LLM writing is homogeneous doesn't hold up. Rather, what's happening is people tend to use default or generic prompts, and therefore receive default or generic results. That's user choice, not a technological constraint.
In short: people were never uniformly smart or hardworking, so blaming LLMs entirely for declining intellectual rigor is oversimplified. The style complaint? Also overstated: LLMs can easily provide rich diversity if prompted correctly. It's all about how they're used, just like any other powerful tool in history, and just like my comment here.