Thoughts on thinking

> But now, when my brain spontaneously forms a tiny sliver of a potentially interesting concept or idea, I can just shove a few sloppy words into a prompt and almost instantly get a fully reasoned, researched, and completed thought. Minimal organic thinking required.

No offense, but I've found that AI outputs very polished but very average work. If I am working on something more original, it is hard to get AI to output reasoning about it without heavy explanation and guidance. And even then, it will "revert to the mean" and stumble back into a rut of familiar concepts after a few prompts. Guiding it back onto the original idea repeatedly quickly uses up context.

If an AI is able to take a sliver of an idea and output something very polished from it, then it probably wasn't that original in the first place.

Familiar scenario. One thing that would help is to use AI as someone you are having a conversation with.

(most AIs need to be explicitly told before you start this). You tell them not to agree with you, to ask more questions instead of providing the answers, to offer justifications and background as to why those questions are being asked. This helps you refine your ideas more, understand the blind spots, and explore different perspectives. Yes, an LLM can refine the idea for you, especially if something like that is already explored. It can also be the brainstorming accessory who helps you to think harder. Come up with new ideas. The key is to be intentional about which way you want it. I once made Claude roleplay as a busy exec who would not be interested in my offering until i refined it 7 times (and it kept offering reasons as to why an AI exec would or would not read it).

Yeah I haven't found the right language for this yet, but it's something like: I'm happy and optimistic about LLMs when I'm the one doing something, and more anxious about them when I'm supporting someone else in doing something. Or: It makes me more excited to focus on ends, and less excited to focus on means.

Like, in the recent past, someone who wanted to achieve some goal with software would either need to learn a bunch of stuff about software development, or would need to hire someone like me to bring their idea to life. But now, they can get a lot further on their own, with the support of these new tools.

I think that's good, but it's also nerve-wracking from an employment perspective. But my ultimate conclusion is that I want to work closer to the ends rather than the means.

You are in a maze of twisty little passages that are not the same. You have to figure out which ones are not.

Try this. The world is infinitely complex. AI is very good at dealing with the things it knows and can know. It can write more accurately than I can, spell better. It just takes stuff I do and learns from my mistakes. I'm good with that. But here is something to ask AI:

"Name three things you cannot think about because of the language you use?"

Or "why do people cook curds when making cheese."

Or how about this:

"Name three things you cannot think about because of the language you use?"

AI is at least to some degree an artificial regurgitarian. It can tell you about things that have been thought. Cool. But here is a question for you. Are there things that you can think about that have not been thought about before or that have been thought about incorrectly?

The reason people cook curds is because the goal of cheese making was (in the past) to preserve milk, not to make cheese.

Not all processes are the same, though. I strongly suspect any efficiency improvements will come in processes that didn't require much "thinking" to begin with. I use it daily, but mostly as a way to essentially type faster—I can read much faster than I can type, so I mostly validate and correct the autocomplete. All of my efforts to get it to produce trustworthy output beyond this seem to trail behind the efficiency of just searching the internet.

Granted, I'm blessed to not have much busywork; if I need to produce corporate docs or listicles AI would be a massive boon. But I also suspect AI will be used to digest these things back into small bullet points.