# AI In A Nutshell

There are lots of fantasies about AI. Both utopian and dystopian.

But an AI is in fact just an automated statistical machine. "AI"
is just a name given to it. So please stop saying buzzwords.

And this is AI in a nutshell:

Suppose that

f(0) = 1
f(1) = 3
f(2) = 2
f(3) = 5

A computer can calculate a reconstruction of f(x), F(x) = 1.1*x +
1.1 which is the optimal line that minimises the sum of squared
residuals (F(x)-f(x))^2. To make F(x) close to the actual f(x).

Or other F(x)s based on other loss functions (ways of how to measure
error).

Because we don't know the f(x) other than those few points. In fact
everyone is doing this during experiments. This is basic statistics
for experiments.

Then people say that the computer "learned" the data set. And F(x)
is the "studied" result.

"Neural Networks" are just some types of computer memory to store
mathematical functions with their parameters.

And when you chat with an AI, you are just walking around on F(x).
It is just that for those sophisticated AIs, the data points are
hundreds of billions many and F(x) is not a simple line. But the
fundamental mechanisms are exactly the same.

So AI consists of two things.

1. How to make data mathematically. Like making similar words
mathematically similar. So that "King" - "Man" + "Woman" = "Queen"
holds mathematically.

2. How to find F(x).

In fact lots of power comes from 1 rather than AI itself.