Ignorance is My Superpower

By Barry Rosenberg, Technical Writer

As a technical writer, I inhabit two types of situations:

  • I am new to a technology.
  • I become involved with a technology long enough to develop some expertise with it.

Expertise gets glory; ignorance gets disdain. Experts swim through technologies like an olympic swimmer attacking the 200 medley. Newcomers thrash around in underground wells.

But ignorance has a sneaky power. All that thrashing makes you not only notice the water but also understand why others are so afraid of it.

In late 2015, my manager asked me to lead development of a course that introduced machine learning to Google’s engineers. I knew nothing about machine learning. Until then, I was accustomed to learning new programming languages and APIs and the bread-and-butter technologies that technical writers work on. Machine learning turned out to be an entirely different programming paradigm. The more I read about it, the more puzzling it seemed. “Wait a minute — you just feed data into a machine learning program and it learns how to make predictions?” Machine learning sounded delusional and anthropomorphic. It also sounded kind of magical. I was determined to figure out the trick.

I started by asking questions. The engineers and scientists at Google tend to be rather accomplished. So, my questions were the equivalent of asking Einstein, “I’ve been hearing great things about this E = mc² thing, Al, but what is E, what is m, what is c, and is that tiny number above the c a typo?”

If I were an expert, would I have bothered doing that? No. I would have just assumed that our students already knew what E, m, and c were. Being ignorant, I asked the really dumb questions about loss and regularization and overfitting. While thrashing, I wrote down the answers in a way that made sense to my ignorant mind. Linear regression, for example, reminded me of graphing cricket chirps against temperature.

I found myself repeatedly dragged under by wave after wave of ML jargon. Eventually, I convened a group of experts to create an ML Glossary for newcomers. The group argued over almost every definition. My side of the argument was typically, “I’m sorry, but I don’t understand that definition.” We refined extensively, searching for that slim intersection of accuracy and clarity.

My notes, along with the help of other technical writers, instructional designers, and — yes, several bona fide world-class machine learning experts — formed something called Machine Learning Crash Course. Eighteen thousand Google engineers took the course to get started with machine learning. They liked it, so we decided to offer this course to everyone, everywhere, for free.

I hope you enjoy the course and the glossary. Benefit from my ignorance. I’ve already asked the dumb questions, so you can reap the benefits.