[Youden]

> Problem with this is that facial recognition is not perfect.

Why is that a problem? Are there any investigative tools that are perfect or is there a reason why facial recognition should be held to a higher standard?

> It's no better than fnger printing which often leads to unjust arrests.

You need to be more specific. Why is it no better? Has facial recognition been demonstrated to lead to more unjust arrests than other investigative methods?

> There is no replacement for detective work and the government shouldn't be lazy to exclude it.

Who's saying that facial recognition is a replacement for detective work? It's just another investigative method, like looking up a license plate or asking people at the crime scene what they saw.

> The best it can do is narrow down suspects.

It can also help find suspects when you don't have any other leads. Why isn't that good enough?

[jackson1442]

The problems come from when it is hailed as perfect and arrests are made solely on a facial recognition result[0].

Facial recognition has also time and time again proven to be racially biased[1][2].

Not to mention how easy it is to create a surveillance state with facial recognition[3].

[0] https://threatpost.com/lawsuit-claims-flawed-facial-recognit...

[1] https://www.cnn.com/2019/12/19/tech/facial-recognition-study...

[2] http://sitn.hms.harvard.edu/flash/2020/racial-discrimination...

[3] https://www.washingtontimes.com/news/2019/dec/9/social-credi...

[Youden]

> The problems come from when it is hailed as perfect and arrests are made solely on a facial recognition result[0].

But is facial recognition itself the problem here? It seems to me the problem is the human who makes a decision based on flawed evidence. This is likely to happen based on other investigative methods as well and not just facial recognition (e.g. "the old lady across the street was sure she saw you breaking into that house the other day").

> Facial recognition has also time and time again proven to be racially biased[1][2].

If we know about the bias, we can correct for it. First in training and decision making and then through improving the facial recognition models.

And again, I'm sure there's racial bias in other investigative methods as well.

> Not to mention how easy it is to create a surveillance state with facial recognition[3].

So long as you have the ability to install millions/billions of cameras throughout the state and put them under centralized control. If that's legal in the US, the problem is that the law allows it, not facial recognition. In most European countries, such a thing would be incredibly illegal.

[heyoni]

I seriously doubt anyone is being convicted based on that alone. Look, once you suspect someone was at the riot, you can then start to look for other things like, license plates coming in and out of DC, or beacon data or phone tower records. Some of these people were shouting their names out to cameras along with the city they were from.

Facial recognition is like watching footage from a bank robbery, and then recognizing the person in it, except a computer does the initial work and a human being verified it before making any moves. I’d be worried if high def cameras were on every corner but this footage was taken at the scene of a crime, by reporters and criminals alike. So what if a private group runs it through a filter?

[Master_Odin]

At least in this instance, the racial biases of facial recognition will not be a problem for the overwhelming white mob.

[DangerousPie]

Nobody is suggesting arresting anyone solely based on facial recognition.

[jackson1442]

Nobody suggested it, but it's still happened. Here's a better link than what I posted above: https://www.nytimes.com/2020/12/29/technology/facial-recogni...

[t-writescode]

Have you seen the sort of things that get people thrown away in prison for murder for decades? Absolutely flimsy evidence like an insane prosecutor that thinks some guy can go to a ball game and flee as fast as possible to go kill a person and then come back?!

Yeah... I don't trust that.