@tjrobertson52

Shortly after ChatGPT came out, there was this idea that pretty soon it's not going to matter how intelligent you are because AI is effectively going to scale us all to the same intelligence level. And this made sense to me, maybe not ChatGPT specifically, but it does seem like eventually AI will be as smart as any human. And since we all have access to the AI, we'll all be equally smart, right?

But I think the opposite is actually happening. Cause the thing is, we as the human are still the boss, at least for now. And AI is like the talented employee that never really takes initiative.

Like they'll do exactly what you ask them to do, but nothing else. And also they don't really see the bigger picture or have any context about what you're asking them to do. And I think what we're finding is that human intelligence is not only important, it's more important, more specifically, the ability to understand and articulate complex ideas.

Just in the last few months, these LMS have been able to handle way more context and multi-step reasoning within a single prompt. So the biggest limiting factor in these LMS performance is your ability to provide the necessary context and explain very specifically and clearly all the steps involved in the process. These days, it's not uncommon for me to provide 12 different documents and 500 words of instruction on the very first prompt.

A year ago, most large language models would have been confused by this. The more context and steps you provided, the less likely it was that they would actually follow each specific step. Nowadays, especially Claude can handle tens of thousands of words and still follow my instructions very closely.

Of course, these large language models still have a long ways to go. A lot of things they're not all that great at. But for anyone with a deep understanding of their specific skill or industry, it seems to being able to articulate that to an L O.

M is going to be a very valuable skill for the foreseeable future.