Hacker News Reader

Comment by kortex | Hacker News Reader

It seems unclear whether face detection unto itself is allowed. Eg automated face blurring. The spirit is definitely the one-to-many searching.

This law could certainly use more clarity.

View on HN

I think it's good that it is simple. People are constantly complaining about the excess of legalese in various contexts. A law that can be understood is cool and a breath of fresh air. I doubt the judges will have any trouble parsing between face blurring and face recognition.

Also, the definition excludes the face blurring use case. "'Face Recognition' means the automated searching for a reference image in an image repository." Although that detail does mean that if you're not searching for "a reference image" then you would technically be clear of the law's restriction. What if you build features based on a reference image then delete the image? Or what if what you're searching for is a person and the image is not the search target?

That I suspect would take more effort to parse. But the penalties are pretty light and the provable damages done by facial recognition are pretty minor, so if organizations get a lot of value out of facial recognition then they might want to try rolling the dice.

I think in a way you are countering your first point that it's good that it's simple. I'm not big into the facial recognition space but I'd imagine that the tech is saving personalized feature vectors and not images. Thus making the definition already moot. Then again maybe that means the language isn't simple enough.

That's a fair point but most systems capture the face chips for human comparison. Is it still capturing "features of images in a repository" if your face descriptors are anonymous with no face images? This is imho a loophole they missed

Although that detail does mean that if you're not searching for "a reference image" then you would technically be clear of the law's restriction. What if you build features based on a reference image then delete the image? Or what if what you're searching for is a person and the image is not the search target?

Any reasonable interpertation would find that the representation used during search doesn't affect what is happening.

Even if the representation isn't an image? The code seems pretty explicit on that point. The courts seem to only do interpretation where the code is ambiguous. They don't seem to guess where the law is explicit. That being said, if there is a problem they can always just change the law. Maybe the signalling effect will be enough. Companies won't want to adopt expensive technological workarounds while knowing that the city is motivated to close loopholes.

Yes absolutely.

Searching for text uses a TF-IDF or BM25 representation.

The only way to make image search work is to use a non-pixel representation. No court will find you aren't searching for an image because of that.

Sure, you have to use a non-pixel representation to make an image search index. I understand that. However, if the end result of the search is not an image, then is it the same? For example, you could learn some very basic features from an image, like "person's skin has this hue, their eyes are this color, etc." If you wrote that down and made it searchable, with the search target being a person's unique identifier (SSN or whatever), it would surprise me if that ran afoul of this rule, even if the features were extracted from an image. You can't be searching for a reference image if there is no reference image to search for.

Maybe I'm wrong though. IANAL and honestly it doesn't matter. But if I were writing this law, I would write that no one would be allowed to use any system that extracts and searches using or makes searchable information about human faces. That would seem to cover my proposed workaround, and I doubt it would hurt any other use case that they meant to protect.

It seems unclear whether face detection unto itself is allowed. Eg automated face blurring.

It's pretty clear that isn't banned. See point 6 in the explanation:

6.Face Recognition means the automated searching for a reference image in an image repository by comparing the facial features of a probe image with the features of images contained in an image repository (one-to-many search). A Face Recognition search will typically result in one or more most likely candidates—or candidate images—ranked by computer-evaluated similarity or will return a negative result.

Face detection isn't face recognition in that case, I believe. Face detection is identifying a face in general, not any specific face. Face recognition is finding a specific face in a set of images.

Yes - we seem to be in adamant agreement? (To quote myself: "It's pretty clear that isn't banned.")

It is written to outlaw recognition (ie, the search action), NOT detection.

Oh, yeah, I misread that as "is banned" for some reason. My mistake, sorry for the confusion.

Reply on news.ycombinator.com