[webmaven]

I'll take a crack at an explanation:

Google has machine learning technology that automatically reads street address numbers on houses, regardless of angle, color, size, focus, or typeface.

Using the same type of algorithm to answer "does this ad have anything that looks like a button in it?" is relatively simple by at least two orders of magnitude.

So rather than it being "easy" in some absolute sense, what's meant here is that it is easy compared to similar tasks that we already know Google does routinely and has largely automated.

[yuriks]

Lots of things have button and aren't maliciously masquerading as download links. Furthermore, if they start doing that then ad producers will start changing their ad images to evade the algorithm. It's much harder to use ML for a problem when you have a malicious opponent actively working against it.

[SmellyGeekBoy]

It should be as simple as:

- Is this a download page?

- Does this ad contain a button-like image with the word "Download" on it?

Presumably the people creating these ads are specifically targeting pages with the word "download" in to create this confusion, it doesn't seem like rocket science for Google to apply the exact same rules.

[wordupmaking]

The easiest thing is to say something is too hard to even try that you don't even want to try for other reasons, such as green paper.

And changing the images to evade the algorithm would also directly affect how likely they are to trick people, so it's win/win?

[Sephr]

ML could be used as a tool to empower the reviewers to to review the most suspect ads first.

[rvnx]

There is a human behind every single creative shown on Google AdSense. There are even rules around these creatives (need to have distinct border and need to say the name of the advertised product).

[webmaven]

Yes, but evading the filter by making the ad not look like a download button (or link, I guess) also solves the problem.