Threshold Numbers and Us

Generative AI have threshold numbers.

A Generative AI model trained on 10 or 100 pieces of information cannot do much, but after a certain threshold it can instantly create anything from a quantum mechanics explainer to a video of twenty ducks wearing bibs with glockenspiel background music.

It seems clear that we also have threshold numbers. There are only so many ways anyone can mix 10 or 100 bits of information together intelligently to create something new, but once you get to thousands, tens of thousands, or millions of pieces of information, that person might be Da Vinci.

Whatever that number is, it must be contextual: the more general and complex the context, the more information needed. A 10,000-piece puzzle is more complicated than a 100-piece puzzle. Different contexts have different numbers.

And whatever the number is, it must remain difficult to pinpoint even after we identify the context. Knowing how to bake a sheet cake might be enough to say that you can make a two-layer wedding cake, or it might not. The only real test is to get into the kitchen, bake a two-layer wedding cake, and see if it does the job.

If there is any lesson to this for those of us in the working world, it is to keep moving and growing because no one can ever accurately say when we have learned enough. The world changes daily, life is school, and school is always in session. Like the title of former PayPal CEO Dan Schulman's podcast, "Never stand still."

Keep collecting information because you never know which next piece could change your life or someone else's.