Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A man spent a year in jail on murder charge that hinged on disputed AI evidence

theregister.com

Post reply on HN
Print view View original

141–150 of 211 posts

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#141

Post by arp242 »

dragonwriter wrote: post #127
arp242 wrote: post #84

Earlier quoted context omitted.

> Fingerprints are directly-recorded facts. That's not really how it works; false positives exist, especially if you consider that a lot of times you have incomplete or smudged fingerprints and there is actually some amount of subjectivity involved in determining a match. See e.g. [1]: "only two properly designed studies of latent fingerprint analysis had been conducted. These both found the rate of false matches (kn…

> That's not really how it works Yes, it is. > false positives exist Matches (positive or negative) are not the same thing as fingerprints. Fingerprints are directly recorded facts. Matches are matters of interpretation (the fact of a match reported by a particular system is a distinct direct fact from the fingerprint itself, but again the significance is...what expert testimony exists to establish or challenge.)

The reality is that when "fingerprint evidence" is brought to court that it's not foolproof and has a sizeable margin of error that should be accounted for when considering the verdict, something that is currently often ignored. You can argue semantics all day, but that's the way things work right now. This seems like an exceedingly pedantic point to make.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#142

Post by giantg2 »

tooltower wrote: post #114
giantg2 wrote: post #87

Earlier quoted context omitted.

That doesn't exactly support your claims for most of forensics. If you read/skim the NAS report these articles are based on, you will see that much of the issue is not with the underlying science, but rather in the "interpretation" part. Yes, the odontology seems to be unsupported. But even the fire analysis that you list as an example is said to have a scientific basis like chemistry, but that the interpretation par…

While my stance on forensics is not as negative as the parent comment, I don't see how you are drawing such a sharp distinction between the "science" and the "interpretation". They are not separate things. When a scientist proposes a new test, they include measurement techniques as well as a method of interpretation. If a medical test is used in a way that is frequently prone to misinterpretation, it would be fair to…

"If a medical test is used in a way that is frequently prone to misinterpretation, it would be fair to call that entire test useless."

And yet there are many examples of this with malpractice, unnecessary deaths, etc. Lyme tests are notorious for false negatives depending on the lab you use. It's also a widely held belief that you want tests done in the middle of the week instead of on a Friday or weekend because some techs just want to get out of there and are more prone to mistakes.

The reason I'm drawing a hard line is because of fact vs opinion. This same line gets drawn in court. They will ask experts if it's a fact or an opinion that they testified about (usually to attack a witness). Autopsies are full of facts, and then there are opinions on what those facts mean. If it's an opinion, it's possible to counter it with another professional opinion.

The only sort of opinion they should be sharing is included, excluded, or undetermined. Hair is not a "match", but it could include or exclude someone and maybe mathematical probabilities.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#143

Post by z3ncyberpunk »

giantg2 wrote: post #93
BoorishBears wrote: post #89

Earlier quoted context omitted.

Because other surfaces are not a purpose built sensor designed to detect the features of a finger then repeatedly trained with a specific set of finger scans? Like are you joking? Saying TouchID validates fingerprinting is like saying MRI machines validate psychics.

Are you joking? It seems you two have a different idea of what is being discussed. The underlying fundamentals of finger prints are solid. The part that can make them ineffective is when they are using insufficiently tested tools or matching on too few points. And of course misrepresentating what a match actually means and how it pertains to the case.

I can effectively burn off my fingerprints completely... They are not completely effective at all

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#144

Post by ransom1538 »

x86_64Ubuntu wrote: post #137
leephillips wrote: post #38

This year in jail was awaiting the “speedy trial” that is supposed to be guaranteed by the US Constitution. I think that in itself is as much of an issue as the quality of the evidence. How can the government justify taking year away from the life of a man who has not been convicted of anything?

A lot of times, people have to waive their right to a speedy trial. Apparently, in some districts they don't like it when the accused doesn't waive their right, so much that the defense attorney will get assaulted by the judge. https://youtu.be/jAv1QUkHFSY?t=397

This. "Michael Williams, 65, who denied any wrongdoing, sat in jail for 11 months awaiting trial for allegedly killing Safarian Herring, 25."

Do not waive a speedy trial ever. The judge will want you to, your attorney will want you to ($), the DA will want you to. Tell them to go fuck themselves.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#145

Post by backtoyoujim »

>“The problem? When we made it possible for anyone to generate art with artificial intelligence, barely anyone used it to make actual art. Instead, our AI model was forced to make videos for random inputs, trolling queries, and NSFW intents.”

I am struggling with the use of the word "forced"; the researchers put their software out there to be used by anyone with an internet connection. The researchers created that dynamic.

And how are they not seeing that AI created no-humans-used-to-make-this-porn could be a net good.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#146

Post by ransom1538 »

sdenton4 wrote: post #103
theelous3 wrote: post #92

Earlier quoted context omitted.

> Most forensic science is junk. I'm ignorant on the topic. Got links?

Forensics is full of gee-whiz-cool csi stuff which eventually is shown to be worthless. Polygraphs are a great example; they're still very wisely used in the US, despite piles of evidence showing they're prone to both false positives and false negatives. Sibling comments saying they aren't forensics because they've been discredited are missing the point: these are typical crime tech, and have been lately discredited,…

https://innocenceproject.org/what-is-bite-mark-evidence-fore...

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#147

Post by tzs »

bagels wrote: post #126
tzs wrote: post #80

Earlier quoted context omitted.

> Fingerprinting is junk science. How then do you explain the success of Touch ID on Apple devices and the equivalent systems on other phones and tablets? Random people can't just walk up and unlock a Touch ID device with their fingerprint, which suggests that fingerprints are in fact a very good way to tell people apart. Yes, it is possible to make a fake fingerprint that Touch ID will not be able to distinguish fro…

How many people have you let try the unlock your device with their fingerprint? Tens or hundreds of Millions? That'd be a closer comparison for how a search of fingerprint database work.

Searching a big DB is one way to use fingerprints. It is a crappy way, though.

Another way fingerprints are used is to compare fingerprints found at a crime scene to people who have been identified using other methods that have nothing to do with fingerprints.

E.g., if you have someone murdered in his office when working late at night, and security cameras show that five other people were in the building at the time of the murder, and you are able to get a good set of fingerprints off the murder weapon you don't need a database. You take the fingerprints from all five of those other people, and if one is a very good match and four do not match, you concentrate most of your effort on the one that matched.

It's actually quite similar to how DNA evidence has been used and misused. I don't know how they compare DNA nowadays, but when DNA evidence was first making waves getting people convicted they only compared two DNA samples at a few points. For a given sample there would likely be several people in the country that matched.

That's fine if used right, like fingerprints in the earlier murder hypothetical. Narrow it down to only 5 people who could have committed the crime, get a DNA sample from the crime scene that must be from the criminal, and if that matches exactly one of those 5 suspects it is strong evidence they are the criminal.

Have no suspects yet, run that same sample through a database, get exactly one match, and conclude that must be the criminal. Totally bogus. A lot of people were convicted in the early days of DNA matching that way.

The database method can be made sound, but only if the database includes everybody. Match against a database that includes everyone, only get one hit, and you've probably got your criminal. But if the database includes everyone you are likely to get several matches.

As I said, that was how it used to be. I know DNA sequencing has gotten faster and cheaper over the years, but I have no idea if routine forensic DNA matching now matches enough to make matches unique except in the case of identical siblings.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#148

Post by aeturnum »

phsau wrote: post #134
aeturnum wrote: post #101

Earlier quoted context omitted.

Just fyi - Police are generally quite well paid for the area they live in. For example, in Alameda (the county I live in) it's common for police to make over $450,000 a year[1]. Also, while police to have a more-dangerous-than-average job[2] (and face unique risks - they are probably the most likely profession to die of a gunshot on the job), there are much more dangerous jobs[3]. So if you have been excusing police…

Where in source [1] does it show that it's common for police to make over $450k a year? The 'Avg Total pay & benefits' sorted by title has Chief Information Officer/Registrar of Voters as the entry into that level of remuneration. Police specific above that you only have the Sherrif, which is an elected official. Police pay for the vast majority of roles is significantly below $450k.

That's a good critique! Looking at historical data it seems like "commonly above $300,000" would be more accurate. It's worth saying that listed police salaries are far below these numbers and the officers get there through overtime and other extra payments. Thus the use of 'common' instead of average because you have to look back at what people were actually paid.

Edit: I can no longer edit my original post so this will have to do.

https://transparentcalifornia.com/salaries/search/?a=alameda...

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#149

Post by abrichr »

jes wrote: post #18
tyingq wrote: post #8

It's interesting to me that there's no working process to introduce/accept new types of science in court. Polygraphs aren't allowed (though the police still find ways to use it to bully people). So, there's some process. But you still see stories about subsets of fiber, hair, arson, fingerprint, dna, etc, science, that shouldn't have been brought to court.

The legal system will, over time, also have to accept and adapt to the idea that free will is an illusion. (See “neuro-law” for more information on this.)

Link: https://en.wikipedia.org/wiki/Neurolaw

> The rapid growth of functional magnetic resonance imaging (fMRI) research has led to new insights on neuroanatomical structure and function, which has led to a greater understanding of human behavior and cognition. As a response, there has been an emergence of questions regarding how these findings can be applied to criminology and legal processes.

I hope they have a more solid foundation of evidence than fMRI.

Eg from https://www.fastcompany.com/90520750/duke-university-researc... :

> The researchers reexamined 56 peer-reviewed, published papers that conducted 90 fMRI experiments, some by leaders in the field, and also looked at the results of so-called “test/retest” fMRIs, where 65 subjects were asked to do the same tasks months apart. They found that of seven measures of brain function, none had consistent readings.

Also https://www.pnas.org/content/113/28/7900 :

> Functional MRI (fMRI) is 25 years old, yet surprisingly its most common statistical methods have not been validated using real data. Here, we used resting-state fMRI data from 499 healthy controls to conduct 3 million task group analyses. Using this null data with different experimental designs, we estimate the incidence of significant results. In theory, we should find 5% false positives (for a significance threshold of 5%), but instead we found that the most common software packages for fMRI analysis (SPM, FSL, AFNI) can result in false-positive rates of up to 70%. These results question the validity of a number of fMRI studies and may have a large impact on the interpretation of weakly significant neuroimaging results.

Re: A man spent a year in jail on murder charge that hinged on disputed AI evidence

#150

Post by FireBeyond »

PragmaticPulp wrote: post #3

The article says that employees of the AI company (ShotSpotter) manually reviewed and classified the sounds as gunshots: > records showed that ShotSpotter actually initially picked up what sounded like a firework a mile away, and this was later reclassified by ShotSpotter staff to be a gunshot at the intersection where and when Williams was seen on camera. So the AI didn’t even make the call. The staff did, manually.…

What almost invariably happens in cases like this or the cell phone surveillance boxes (Stingrays) is that if the underlying technology or behavior is challenged or questioned strongly enough, the evidence introduced by them, if not the case itself, will be withdrawn, in order to try to avoid any negative precedent against the device or technology being set.
Post reply on HN