[flagged]
What you've probably just never noticed is that voice is a critical part of communication in written form just as much as in speech. And this is especially true when communicating in a crowded public environment like the internet.
Voice helps the receiver cohere your message as a distinct signal to pay attention to and draws it out from the noise of competing communication in their day, on their screen, etc. It helps them hear you -- and what you mean to say -- in your prose. It also gives the receiver subtle but meaningful insight into where you're thoughts are coming from culturally, experientially, and ideologically as your voice leaks that background information about you as subtext.
Because LLM's are a compression/decompression technology, letting you draw a lot of content out from a small leading prompt, they infill that output content with a voice that has characteristic and common phrasing, characteristic and common tics, characteristic and common structures, etc. So output from any given LLM, and from most LLM's trained in similar ways, ends up tending towards the same voice. i.e. they just to contribute to the noise.
It's deceptive, because your English teacher probably would grade their prose more highly than yours and so of course there's a part of you that thinks that's what you want. But impressing your English teacher is only part of communicating. The other part is making yourself heard.
LLM's make your prose easier to analyze or evaluate but much harder to hear.
Writing well is not easy. PG has written many essays about this very topic.
LLMs do not write well.
> Strategic thinking? The LLM has more context than you. Sure, you know more about your company, but AI knows more about the rest.
This is a mistake, and untrue. The LLM has more general knowledge, yes. But it’s the specific knowledge (aka “context”) that makes all the difference. It’s the same reason management consultants are, by and large, useless; they aren’t dumb, but they lack context.
So no, the LLM doesn’t have more context than you. In fact, its context window is literally limited. You have decades of experience and knowledge and patterns you’ve matched, and the LLM has none of that.
Do not fool yourself into thinking it has more context than you. It doesn’t. It has more knowledge. Those are very different.
(Incidentally, the same is true of humans; the most knowledgeable ones are not always the ones making discoveries or changing the world. Often, it is specific context that matters more, as the application of a solution to a problem requires seeing the right problem. LLMs are a solution engine, but they have no metric for what is a “right problem”)
LLMs can't replace the thinking. Unfortunately, they can produce output that masks a lack of thinking.
As a recipient of AI-enhanced texts I must say: AI is not good at making things clearer. Good communication requires that the sender makes the message as simple as possible. AI is anti-simple, because it likes to produce more words, not less.