Hacker News Reader

Comment by ryanb | Hacker News Reader

Super interesting. A lot of people feel the same way. Why do you think you weren't able to get any traction for this?

They probably look at the metrics and see that a lot of people are going to the Pinterest results. That the results are garbage and undermine the Google images search experience is much harder to measure.

But isn’t this a sign that it’s actually useful for people? At the same time I have to admit that google image search is probably their worst vertical.

I'd say it's more of a sign of what shows up highest in their index rather than any indication on the quality.

If people are actually navigating to the results and not bouncing immediately due to the login wall that is a sign that people are finding them useful.

Is it that weird to believe that lots of people have Pinterest logins?

With image search more than other types of search IMO, because the source of the image is less prominent than the image itself, so ignoring specific domains is harder. I've clicked Pinterest links over and over in google search when on my phone when I had no interest in them, but it's hard to avoid because it isn't always clear what is what.

Welcome to the future.

Everything is decided by AI based on data we don't really understand.

If only there was a way to track the number of seconds they spent on pinterest until the back button was hit and... factor that in as a feature in ML. voila.

I don't think it's that simple. When a user goes back to image search results, is it because they are unsatisfied with the result they clicked on, or are they looking for even more, hoping to compare? And when a user closes out instead of doing back, is it because they got what they wanted or because the results were so garbage they gave up? Based on my own usage of image search results I don't think you would be able to tell when I was happy with the results or when I wasn't. But you can probably tell that my usage of image search went down when it started getting shitty due to this spam.

What are your use-cases for image search? Mine are pretty straightforward; find a higher quality/uncropped version or to find the source (article, recipe, or photo set). Use case #1 I click through save and exit. Use case #2 I engage like I would for web (text) search.

#1 might be confused for not finding the intended result, but could be discerned after looking a bit more.

Also, if a lot of people add in "-site:pinterest.*" (or click on pinterest sites but come back, then find the intended result) that might be a hint it's not adding real value.

I think my use cases are pretty varied, anything from looking for pictures of a place I'm considering going hiking, looking for different views of a product I'm considering purchasing, looking for other examples of an artists art, to trying to figure out where some image originated

This answer is super spot on, and this will be the story of our lives for a while.

It could be better than what we have now, assuming somewhat smart AI. Right now, data gives plausible deniability, and is the universal scapegoat - you get to do whatever you want, backing this up with misapplied A/B tests and telemetry, with zero statistical rigor. And if someone complains about your decisions, you point to these tests and that telemetry, and cry "we're a data driven company, and the data proves we're right".

I mean, it would help to be honest about the metrics that the A/B tests are evaluated against. If you optimize purely for growth and engagement, then your test results will reflect that - even moreso if your tests are scientifically sound.

I'm continually surprised how companies manage to sell "let's try to keep users glued to the screen as long as possible" as doing something for the general good.

I have little hope that the people not understanding their data today and lacking any statistical rigor will be able to train an AI to do better. Figuring out whether an AI gives you garbage results seems to be a harder problem than figuring out whether your own results are garbage.

No one said it had to do with AI. The point being made above is that while you and me may dislike Pinterest, the data may be showing that to most users, it's actually very valuable.

If anything, ML would be able to detect my and your use cases and intelligently hide Pinterest for us while hiding it for others. So chances are they aren't using ML yet which is why it's just favoring the majority rather than us (the minority).

Reply on news.ycombinator.com