In what domain have LLMs been demonstrated to so-called “think” consistently better than a college- or even high-school-educated adult?
I've got dyslexia and adhd that makes it hard for me to do long form software engineering writing like requirements analysis and test plans. With an LLM I can really quickly sketch the use case, create a reasonable list of requirements, break those into stories, and write implementation stubs and unit test cases. It's like having a really decent project manager on the payroll, when before I couldn't manage the complexity of the writing a good system spec.
Obviously in both cases it's me doing the thinking. But again just me, in both cases, would be stuck and completely overwhelmed.
This kind of co-regulation is incredibly valuable, even for me as a fairly educated developer. But perhaps you're right. It's not real thinking. I would say that this kind of AI assisted co-regulatory interaction could be called "co-thinking".
The idea is that the llm has certain cognitive and material weaknesses that I cover for it, like fact checking, big picture thinking, and true identity/agency. And in at the same time, it's able to cover certain cognitive and communication weaknesses that I have.
The result is that I'm much more technically independent than ever before, and can do things in my career that my disabilities prevented before. That matters a lot to me, and is my very personal reason for believing this tech will matter to humanity.