What is AI?

Disclaimer: I am not an AI researcher. I am merely an opinionated computer engineer who trained one or two neural networks during their studies.

Is Artifical Intelligence really intelligence? Does being able to compute 2948×7938 in an instant make someone intelligent? Make something intelligent? This thought is way beyond the scope of this blogpost.

Most of the time, when we talk about "Artifical Intelligence", we talk about artificial neural networks, or derivations of this principle.

What is an artificial neural network?

In essence, neural networks are big mathematical functions. They takes values in, spit values out. Those functions have tons of parameters that do not represent anything meaningful to us. We will have to change those parameters little by little to nudge the neural networks to give the expected output for a given input. And if we multiply that by enough known input-output pairs, we can hope that this function can later be generalised to more input-output pairs that we did not "train" it on.

Because they can take so many values as entry, those functions can be used by developers to solve problems too complex to solve with regular programming. For example : is this a picture of a cat or a dog?

Text based generative AI work the same way. They take an incomplete text as input, and output words to complete said text.