Alex Hormozi

computers, right?

And so fundamentally you train a computer the way you should train a human.

The reality is that most people don't train humans like they should train computers.

And as a result, they're bad at training.

But one of the things that if you've ever followed my channel at all, I'm a big believer in thinking through operations, thinking through observable behaviors, right?

And so this has actually been an amazing translation for my skillset into training AI, because if you take all of the emotional words out of it, right?

Take all of the ephemeral, take all of the intangible,

out of all of your words, charisma, make it lighter, make like all of these words that people use and just say like, what do you want to have happen?

Which most people do not do because they do not define what good looks like.

If you can actually take the time to define what you actually want, rather than expecting the other person to guess and somehow get it right, or expect the agent to guess and get it right, which is really what we're doing, then all of a sudden you'll be able to be so much better as an AI trainer, which is fundamentally where you're going to be.

so that they can actually do the work that you want them to do and do it at 100 times the speed with no complaints and at 100th of the cost.

And so I'll give you an example.

If you're like, hey, I want you, like this is a very simple example because most people can understand this use case, which is like, hey, I want you to write,

some copy for me, right?

So it's like, hey, write this emo copy for me.

If they write copy and it sounds like AI slop, it's usually because you didn't give anything to it besides write words that are English and correct and make it sound like the internet, which is fundamentally what AI sounds like.

It was trained on the internet, right?

And so it's not the best writing.

But if you say, hey, here's 12 rules that you can never break, and here's 16 writing samples of mine, and I want you to write only according to this, you're going to probably get an output that's probably like five times as good at that.

Now, if you repeat that loop 100 more times, all of a sudden, you'll have an output that's perfectly trained on the patterns, except with a person, they forget some of the things you said 16 times ago, and it takes them time to go type it or time to go learn that feedback cycle.