Learning to play

Learning to GPT is a high leverage thing you can do right now. Like learning how to know how to use AI for a purpose of sorts or for some sort of benefit. 99% of the world does not know how to do that yet. But imagine the productivity or fun gains you can get out of using AI — imagine them to be similar to the gains you get out of using electricity or using the internet (noting that those things also have costs).

I mean I think it’ll be high leverage to know the AI domain — perhaps one day this layer will all be abstracted — and thus I want to write an essay at least to myself to commemorate this gut feeling. When I say learning to GPT, I am surely using the nouns and verbs wrong but I am just commenting more so on the human<>AI potential for interaction, and learning to do that as beautifully as someone knowing how to play the piano.

The piano is honestly a very interesting example. You can model the piano as a piece of technology. On its own, it’s perhaps pretty but makes no noise. And if you don’t know how to play it — as most people in the world do not it — it’s kinda a messy instrument that just creates lots of noise and doesn’t sound very good to the naked ear.

But to those who have learned to play…through practice and training and failure…it can sound like one of the prettiest things in the world.

AI could be a bit like that. As with any tool. The learning to play period may be messy and concerning. It may not be obvious as to why the heck you are playing with this thing in the first place. It is making all of this unwanted noise. It is all over the place. Why? Make the music stop?!

But then you hear a stroke of genius. You see a major gain. You hear that note and your body literally reacts to it.

That is the arc of trying to learn to play.