Christopher Mims

It turns out that the world runs on small models.

We just don't hear about it because it's not as flashy or exciting and it's not a place where companies can make zillions of dollars.

Small models have taken over for translating text and documents into machine-readable text.

Small models are what allow voice transcription on your phone or in any app that does that for you.

Small models are fast.

fast.

Speed is often of the utmost importance.

We still tap on buttons and type on things and ask the cloud to do things for us and expect immediate responses, right?

So a smaller model is what's powering the generative summaries that you're getting on Google these days.

Obviously, they cost less to access and to run.

And then finally, because they can be specialized to individual tasks, they can actually be more capable than large models at things for which they've been fine-tuned.