Hacker News Reader

Comment by rm445 | Hacker News Reader

They had a bunch of CCTV and no name. The AI gave them a few names, they looked at its report which had photos from the guy's social media and were like 'yes that's him', and then got a warrant to go to his house and found the gun.

Perhaps there's some important principle of rights that's eluding me. But it seems like the actual murderer was seen on CCTV and found with the murder weapon, and claiming AI (used as a search tool) as an illegitimate cause for the warrant is a ploy by the defence. Doing their job, sure, but it doesn't seem like natural justice or any broader rights would be served by letting this guy off.

It would be different, the Prosecutor's Fallacy, if the AI->name link was used to justify guilt, but instead the standard of human facial recognition used for getting other warrants from CCTV can be used.

View on HN

> some important principle of rights

Yes, you can't use evidence that was obtained illegally to prosecute someone for a crime. Their guilt or innocence has nothing to do with it.

If illegally acquired evidence was admissible in court, the police would have every incentive to ignore people's rights on a regular basis, conducting illegal searches and seizures, breaking into homes without a warrant, etc., in order to obtain evidence. On balance, this is much worse for society than one guilty person going free.

How is this different from a random passerby identifying the suspect from the grocery store footage, or a police officer already knowing the suspect? They are just clues until any proof is found (ex. a gun that matches the bullets).

Assuming the police didn't lie about the connection and the warrant was issued by a judge, both of those things would be fine. The issue in this case is that the police were not forthcoming about the source of the connection, and the connection itself was from a tool that explicitly says that it is not admissible in court. The warrant was obtained based on false pretenses, making the subsequent search illegal.

A judge probably would not rule in favor of a warrant based solely on a tool whose utility is entirely unproven and which itself warns that it should not be relied upon as evidence, while they might rule in favor of eyewitness testimony.

But a random passerby also isn't admissible in court but could be used as a tip, right? If an anonymous person calls the cops and says "I heard someone say that Bob did it", there's no chance in hell that an anonymous person's hearsay could be admitted as evidence in any court, but if the cops get that tip, decide to check Bob's social media and decide that Bob looks a lot like the guy in their CCTV footage, they're not getting a warrant for a search based solely on the inadmissible hearsay, they're getting it based on their determination that the tip that they got does happen to match the footage they have. If they had to manually make a determination about whether a given person's face matched up with their CCTV footage and the warrant was issued on their determination that it does, it seems very complicated to then figure out to what extent the tip itself has to be admissible as evidence, right? Like, if a psychic called a police department with a tip and the cops put a house under surveillance and the surveillance reveals probable cause to get a warrant, should that kill the case because the tip came from supernatural sources?

Again, if the police have sufficient evidence to convince a judge that a warrant is justified, it doesn't really matter what the evidence is, only that it is not misrepresented to the judge and that the judge deems it sufficient.

The main problem in this case was the fact that the police misrepresented their evidence to the judge. It is theoretically possible that a judge may have issued a warrant on the basis of the facial recognition tool, in which case the evidence from the search would not necessarily need to be thrown out. If that were the case, I'd expect the defense to appeal the validity of the warrant, in which case all the questions you're asking would come into play.

Because for a random passerby you can have them withness and be responsible for the withness statement. Clearview can fill their database with crap, or even fake some data, to manipulate the police investigation. Because Clearview knows this is not only possible, but also likely, the database matches their should not be used as any evidence. If such tools are used, a lot of innocent people are arrested and just arresting you for a murder, even not guilty, will destroy your life.

Eyewitnesses are notoriously reliable as we know

Eyewitnesses you can at least cross-examine on the witness stand.

so, basically - the court wants to treat Clearview's AI as an unreliable witness?

I always hear this argument of "how is this different from XYZ?" when it comes to applying novel technology.

If Clearview AI is indeed functionally equivalent to some random passerby, then what value-add does it actually have over that baseline? Either it's really the same and there's no value to justify deploying it. Or there's something else at play (say, scale) that's worth examining on its merits and risks.

Well there isn't always a random passerby.

Both of those things are legal. That’s the difference.

The cops lied about where the suspect was identified?

So the warrant was granted from false pretense.

Law enforcement use of AI isn't illegal in Ohio, just that software claims to be inadmissable. Hearsay is often inadmissable. But my understanding is that's because it is not reliable, not because it poisons subsequent investigations with illegal forbidden knowledge once a detective hears thirdhand that they should look into someone.

It seems like a catch-22 here to be in that position. They can't cite it for the warrant but if they don't they get accused of being misleading.

It’s not a catch-22. They can use it, and include it in the affidavit for a warrant. They’re just going to ALSO need some other evidence in order to satisfy the judge that the warrant is justified.

It's a catch-22 because the defense can now attack it for being presented in court when the disclaimer says not to.

They apparently did have enough other evidence to satisfy the judge since they got the warrant without it.

> Perhaps there's some important principle of rights that's eluding me.

Convicting the guy is one alternative, but there's another (non-mutually exclusive) option: punish whoever conducted an illegal search. If a warrant legitimised the search, punish the judge instead. If the judge didn't know any better because he was deceived, punish whoever lied to the judge, and so on.

"Might not get a conviction" is a negligible deterrent against police overreach.

> "Might not get a conviction" is a negligible deterrent against police overreach.

But it isn't just that. It's a possible perjury charge for lying to a judge. It's the strong likelihood of ending the detective's career or at least limiting it significantly. It's the political fallout from articles like this one. It's the potential civil rights lawsuit bankrupting against the department and detective. It's the personal shame and guilt that the detective feels for knowing that it's their corner-cutting that let a murder escape justice. Imagine having to face the victim's family if this guy is acquitted...

These are big deterrents.

Then why does this sort of thing keep happening?

This is another reason why police officers should be required to carry malpractice insurance, that they pay for.

How often does it happen? 0.1% of cases? 0.01%?

That figure sounds right - if you're talking about prosecutors being punished.

  A survey conducted by the Innocence Project, Innocence Project New Orleans, Resurrection After Exoneration and the Veritas Initiative looked at five diverse states over a five-year period (2004-2008) and identified 660 cases in which courts found prosecutors committed misconduct, such as tampering with key evidence, withholding evidence from the defendant or coercing a witness to give false testimony. [..] Of the 660 cases examined, only one prosecutor accused of misconduct was disciplined.

https://innocenceproject.org/why-holding-prosecutors-account...

This is how it works, or is supposed to work, in Norway. Illegally gathered evidence is still evidence.

I wonder how well that can possibly work in practice. It can't go over great in front of a jury when someone who is going to be punished for violating the rules talks about the evidence they've obtained, advocating for its legitimacy.

There are no juries in Norway. Evidence is merely a collection of facts, how it was gathered does not alter its factual nature.

I expect the problems are similar in front of a judge.

> Evidence is merely a collection of facts, how it was gathered does not alter its factual nature.

That's obviously not true. The simplest counterexample is a witness report, where the credibility attached to the facts being reported will depend directly on the credibility of the witness. "The suspect verbally admitted he committed the crime while locked up in my car" will mean a lot less coming from a cop who lied to a judge while obtaining the arrest warrant.

That officer is not relating a verifiable fact, it's not evidence. Eye witness testimony is well known to be unreliable. As are confessions. Besides, you've move the goalposts from how to by whom.

> it's not evidence

It’s boring to argue about the definitions of words, but you’re using English and in the countries where English is primarily spoken, testimony is considered evidence.

> found the gun

If any bullets were found and if the gun was shown to be a ballistic "match," the article neglected to mention it.

> Perhaps there's some important principle of rights that's eluding me.

The legal principle involved is called "fruit of the poisonous tree."

https://en.wikipedia.org/wiki/Fruit_of_the_poisonous_tree

> If any bullets were found and if the gun was shown to be a ballistic "match," the article neglected to mention it.

Good point - maybe I read too much into "The search turned up what police say is the murder weapon".

On the principle, I'm contending the tree wasn't poisoned. However they got the guy's name, whether it came out of some high-tech black box, or a detective remembered the guy from somewhere, or they, I don't know, employed a clairvoyant and conducted a seance, once they've looked at the guy's socials and the CCTV and seen it's the same dude, there's no poison.

How do we know the gun found at the suspect’s house was the murder weapon? A gun owned by one person in any given geographical area tends to be owned by at least many others.

And, is it better to let off one hundred guilty people than to convict one innocent.

> and then got a warrant to go to his house and found the gun.

The police claims it's the gun. That has not been proven, and the article doesn't present a shred of evidence.

claiming AI (used as a search tool) as an illegitimate cause for the warrant is a ploy by the defence

The problem here is the same as if the police had coached a witness to identify a pre-selected suspect. The police had already decided that Tolbert was their suspect (albeit without knowing his name) by the time they ran his image through the AI facial recognition, so they discarded all the other matches.

The court also noted that AI facial recognition is no different from an "anonymous informant," which cannot be used to establish probable cause under Ohio law because, very crucially, it can't be questioned under oath.

it doesn't seem like natural justice or any broader rights would be served by letting this guy off

This assumes that the suspect was actually the killer. We don't know if Tolbert was the killer because the police the police conducted a bare-bones sham of an investigation, and lied about how they conducted the investigation in court. They could easily be wrong about the killer's identify, which means they would not only be putting an innocent man in prison, but also that the true killer would still be free and in a position to murder more people.

> But it seems like the actual murderer was seen on CCTV and found with the murder weapon

Hold up, he was found with a gun but it does not say it was the murder weapon. Lots of people in the U.S. own guns. To call it “the murder weapon” requires hard evidence linking it to the murder.

The gun, or a gun?

Reply on news.ycombinator.com