The breakthrough behind ChatGPT

What word comes after Good?

You may think Good Morning, or Good Bye. But you actually wouldn’t say Good Loud. That just doesn’t make sense. For many years, computer scientists have been training AI to unravel this exact problem.

Given a word, our AI predicts the subsequent word. Do that several times, & you’ve generated a sentence.

That is how ChatGPT works.

Trained over all the web, ChatGPT has learned the best way to chat like a human. Nonetheless, this immense feat was only made possible by a breakthrough within the late 2010s. A breakthrough underpinning ChatGPT & perpetually shaping the world we live in.

That is the story of an AI that read & learned from every book, tweet, & website across all the web. And the way it was made possible.

Sentences are long.

After we move beyond a single word, next word prediction is quite a bit harder. Take this instance.

On this context, it is not sensible to say I ate a great Morning. But our AI only looks at good, and spits out morning. Generally even humans need many words to predict the subsequent word. So an AI needs this extra information as well.

Our AI must read many words to predict the subsequent word. ChatGPT can read greater than 8,000 previous words without delay. The natural solution to do this is able to be to feed each word into the AI, one after the other.

That is how AIs worked prior to now. A Recurrent Neural Network (RNN) would take one word at…