About once a month, I’ll do a more-or-less random Internet search for images. The idea is that although most random searches don’t reveal anything extremely useful, every now and then the search results reveal some interesting insights that are useful for machine learning image classification or analysis.

My recent search combined the terms “armed robbery” and a large city that one of my friends moved to recently. In the image below, I blotted out the city name to avoid result-bias.

As is usually the case, the search results didn’t reveal anything unexpected. I was a little bit surprised that most of the results were of people rather than of things like stores or police cars or similar non-people images. The people shown are not labeled as victims or perpetrators, so there’s no way to tell which are which. I did a few follow-up searches for just “robbery”, without the “armed” part and the results were a bit more variable, but were still mostly photos of people.

Well, nothing ventured, nothing learned. I’ll try another semi-random Internet image search next month.

The company I work for encourages a growth mindset — simply meaning a culture of constant learning. I’m not sure what percentage of knowledge gained comes via structured, intentional activities, and what percentage comes via unstructured, ad hoc scenarios. But I’m sure that both channels of learning are important. And I know for sure that learning doesn’t happen by accident — you must continuously explore. Like doing random Internet searches.

I have a copy of the 2006 book that explains what a growth mindset is. I don’t like pop culture science, but this book is quite good and has some solid research basis.