How Does AI Actually Work?

Morning Neural Networks!

Here’s something crazy to think about on this fine Tuesday:

You know all this AI-large-language-model hoopla?

LLM this.

LLM that.

Ya.

So Google basically invented that:

BUT! We kind of have no idea how it works now..

It’s large language-y properties just ~emerged~

Never fear!

Us derds can figure out how AI actually works together <3

So it was 2017.

Google's research team launches this ambitious project:

We’re talking everything on Wikipedia. News articles. ALL the books ever. Forums. Billions of web pages.

And Google’s goal with this is super complicated and difficult to understand:

So they present their new AI architecture...

Bingo!

That’s the foundation of their breakthrough Transformer neural network architecture.

So they repeat this learning process TRILLIONS of times.

For every incorrect answer, the neural network adjusts its connections (weights) a little bit:

For every correct answer, it reinforces those connections:

After processing more text than humans could read in thousands of years, this thing discovers patterns in language we didn't even know existed..

That’s when AI starts to take on a life of its own.

The rest of this story is fricken nails.

Click here to watch the full video 👇

Stay Cute,

Henry & Dylan 🌈

If you’re that sexy friend, subscribe here.