This is a piece I originally posted to my Instagram story in July 2024.

A large online education platform asked me to take on a course. They wanted me to teach deep learning. I thought it over, but in the end I declined. There were two main reasons.

1. A system where the number of examples is what matters

Example-driven teaching is fine in itself. The problem is a system in which the number of examples is the thing that counts. I’ve seen this approach in books with titles like “Master Photoshop in a Month with 40 Examples.” A book like that will inevitably include a chapter such as “How to Make a Neon Sign Effect,” and if you follow along step by step, you’ll be able to produce a neon sign effect. But what happens when you’re handed a different problem? What if you want to use the same features to create not a neon sign but a fire effect? My point isn’t whether you can actually pull off that fire effect. It’s whether this kind of teaching genuinely sparks curiosity, gets you thinking about the root of the problem, and helps you solve it.

Back in my undergraduate days, I ran a programming club. Among the members, some kept studying steadily even when they weren’t showing up to the club, while plenty of others came for a few days and then quit. The group with the widest swing in enthusiasm — the ones who burned hottest and then lost interest the fastest — were precisely the people whose goal was a specific feature. Most of them had no curiosity whatsoever about what lay beneath a given feature or what principles made it work. And the moment they achieved their narrow goal of implementing that one feature, they figured they’d learned it all and stopped coming.

I’m not arguing that example-driven, top-down teaching is bad across the board. Not everyone needs to dig deep. But teaching where only the number of examples matters leaves you with nothing in the long run, for the reasons above.

2. Marketing

This follows on from the first point. “40 examples,” “master it in a month” — claims like these are so sensational and so blatantly far-fetched that they actually irritate me. Does knowing how to do 40 examples really turn someone into a person who can solve genuinely new problems? Can you really master deep learning in a month? In my view, these ads need to be a lot more honest. Why not say something like this instead: “This course will probably help, but if you want to become an expert, be prepared to pour in an enormous amount of time.” That I’d trust more.

Of course, I understand that a company has to make money. But isn’t this taking it too far? Honestly, it’s not just companies — there are plenty of people running these kinds of cons. Everyone talks as if they’re an expert who knows something extraordinary. They say ripe fruit draws the bugs, don’t they. This is a moment when individual discernment matters. Making money is fine, but I’d rather live a little more honestly, saying what I actually want to say. At least for now.