Never mind custody decisions, let's AI up our police cars Not long after the news that UK cops may use artificial intelligence to make decisions on custody, Nvidia was showing off AI-for-cops at its GTC event – except this time it's the vehicular sort. Of course, all of this is par for the course at the GTC event, you’ll always hear about the latest car sensor technology, self- …
at the thought of taking any "AI" assisted desicions through the courts.
If any decision is based upon 'likelyhood' or 'possibility' or 'inference' then it should get thrown out in a flash.
But hey, our PM would never want to introduce this here? Shirly not?
Posting AC just in case.
Presumably the system would alert an officer that the person stood on the corner resembles wanted bad guy Joe Blogs, and then show the officer a mugshot of Blogs from the records. The human officer would be the entity making the decision to question the person on the corner. In this respect using an automated facial recognition system doesn't change anything legally.
I'm assuming that it is currently legitimate for an officer to talk to someone resembling a known wanted criminal based on a photograph or other image.
Where the lawyers might get some business is if this nVidia system, through a fault or bad data, causes an innocent person to be questioned ten times a day.
The optimistic part is the officer overriding the system and not arresting someone for the resemblance or lack there of. They would only have to let the target go once and they'd have that hung around their neck for the rest of their career. No risk for them in a false arrest cause hey it wasn't a false arrest, I had my handy magic box that told me it was okay.
sorry, I meant the criminals will / do adapt, in general, not that every idiot with a "piece" will switch to large-scale financial money laundering. Some "opportunities" will close, others will open ironically, on the wave of techologies implemented to close the old style of crime). Those that don't adapt, disappear over time, in the meantime, the new breed will / does explore those new opportunities. A certain type of crime, e.g. murder by firearm) might decline, but overall, the highly increased monitoring capabilities discussed only apply to a narrow range of "crime".
But hey, cynical mode on, this "street crime" might be a minor fraction, but it's highly emotive and publicized, so... MORE CAMERAS, MORE DRONES!!! And, cousin, business is a-booming!
This technology is going to make those with nothing to hide have something to fear : false positives.
This being the US, imagine a black person walking down the street. This person has no criminal record, but a passing patrol car gets a false positive that he is a criminal wanted by the FBI. Do you really think it will end well for the black guy ?
The car could stop, the officers get out and question the guy and clear the issue by simply controlling his papers or checking the picture of the wanted guy, but somehow I doubt that it will always happen in such a fashion. Some form of brutality is more than likely, and a fatal issue is unfortunately not at all impossible.
A dead innocent because of a false positive - that's not what I call nothing to fear.
There is also the possibility that the system might rule out an innocent person, whereas a human officer might mistake them for a wanted felon.
We humans vary greatly in our ability to recognise faces... some people literally can not recognise their own mother, whilst at the other end of the scale are people whose natural ability means they have long been employed by casinos and intelligence agencies.
There is also the possibility that the system might rule out an innocent person, whereas a human officer might mistake them for a wanted felon.
But there is a variability in human facial recognition which isn't present in duplicate copies of computer software. If one car equipped with facial recognition software decides you are probably Dr Crippen, then all the other ones that drive past you will too.
That's relatively easy to fix if you allow officers to flag false positives and spread that information throughout the system, if officers flag "true" positives the system can also improve by itself.
"We know you're not a wanted felon or anything, but we'd still like to take some photos of you for our records anyway." Or perhaps they won't be asking for consent at all.
In the implementation described, this kit is working with officers, not supplanting them. Today, you could have traffic cops read every number plate manually and compare it to a list... but that process is probably too slow to be useful in real time. It's a job best automated, leaving the officers' eyes available for other tasks.
That said, many people anecdotally talk of how their spelling has deteriorated since they used spellcheckers for most of their writing. Could it be that using a system that flags known bad guys might make human officers less vigilant? I don't know.
The point about intelligence is mis-applied here, because it's not about HUMAN intelligence in tracking crime, it's about information gathering intelligence. There's no argument that any system is incomparably superior to any human in harvesting and cross-checking harvested information against databases. Yes, there's a chance of mis-match, yes, a chance to hack the database, etc, etc. but overall, the systems v. humans is a lost war - in some areas. And the systems only get better, and get better - faster. And those at the receiving end of the system imperfection, well, too bad (for them).
It's the same in other areas - human call handlers are better at routing calls to the right department, but training and employment is costly, so businesses decided, f... quality and f... customer satisfaction of every customer, we'll handle a lot more calls with the new system and overall the satisfaction will improve. And we WILL use that (crappy) voice recognition system everybody hates so much, cause it's gonna save us x amount which we're gonna award ourselves. And it appears that they were right, because we have accepted the inability to get through to ANY right person on the phone, and stopped calling, started e-mailing, then started tearing hair over their useless FAQs, but hey, now we have their useless "AI" chatbots.
The same can be said about translation: f... quality, we're gonna use google translate on our webiste and so whit of 10 people die lafin. 1000 get the massage, more or les.
...is what is said of "military intelligence", and law enforcement doesn't have a much better reputation and is often described as being - let's say - rather challenged on the intelligence front.
NVidia may think that a few artificial neurons can't hurt and may contribute to raise the force's IQ by a few most needed points?
The closing paragraph of the article says I think it’s undeniable that technology like this will lower crime rates.
"Lower crime rates" means reducing the number of crimes recorded. Technology like this might increase arrests or prosecution rates in certain areas of fairly low hanging fruit, but it is far from obvious that it will lower crime rates. The Police budget in the UK isn't going up, and this equipment doesn't look cheap. The money for expensive equipment has to come from somewhere. Would they "streamline" the number of policemen or perhaps take the money from crime prevention, training or just close a few more police stations?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
