Hertz' AI System That Scans for "Damage" on Rental Cars Is Turning Into an Epic Disaster
Technology
163
Beiträge
92
Kommentatoren
2.5k
Aufrufe
-
Use Turo. You can rent basic or fun/interesting cars directly from the owners.
Oh yay another bullshit thing. No thanks.
-
I use an app called GoMore in some places in Europe that allows you to rent cars from other peers. The rental process is cheaper and faster--everything is done through the app--and you avoid these shady corpo practices.
For now till the shit that happens with Airbnb happens there. With the corporations just renting all the cars.
-
Sometimes it's made with lower quality ingredients at the same factory, sometimes it's equivalent.
Some might be lower but some would be the same.
-
I think it's generally a brilliant solution but there are a couple of problems here:
- The scanner seems to flag fucking everything and charge for minor damage where a human would probably flag it as wear.
- No one is allowed to correct the scanner:
Perturbed by the apparent mistake, the user tried to speak to employees and managers at the Hertz counter, but none were able to help, and all "pointed fingers at the 'AI scanner.'" They were told to contact customer support — but even that proved futile after representatives claimed they "can’t do anything."
Sounds to me like they're just trying to replace those employees. That's why they won't let them interfere.
You are spot on here. AI is great for sensitivity (noticing potential issues), but terrible for specivity (giving many false positives).
The issue is how AI is used, not the AI itself. They don't have a human in the checking process. They should use AI scanner to check the car. If it's fine, then you have saved the employee from manually checking, which is a time-consuming process and prone to error.
If the AI spots something, then get an employee to look at the issues highlighted. If it's just a water drop or other false positive, then it should be a one click 'ignore', and the customer goes on their way without charge. If it is genuine, then show the evidence to the customer and discuss charges in person. Company still saves time over a manual check and has much improved accuracy and evidence collection.
They are being greedy by trying to eliminate the employee altogether. This probably doesn't actually save any money, if anything it costs more in dealing with complaints, not to mention the loss of sales due to building a poor image.
-
Too many people these days don't use or have access to credit cards for services like this. Many people I know only use bank debit cards, or worse, use the debit preloaded cash cards issued by their employers' payroll service provider.
Credit cards motivate banks to help you, because if you won't pay, and the business doesn't pay, the bank has to take the hit.
Debit cards will work as well if your bank values it's reputation - but not all banks do.
And I would not trust a preloaded card provider to assist. You are neither their business partner nor their customer and that puts your interests at the bottom of a very long list. You have to hope some law is on your side or that your issue is so trivial that resolving it is more cost effective then dealing with you.
in this case, hertz doesn't rent to who doesn't have a credit card
debit = no rent
-
Virtually any place that accepts a credit card will accept debit cards, too. Actually, most debit cards can be processed as credit cards. The comment you responded to simply highlighted that this trick is much easier to pull with credit card than a debit card, as the creditor hasn't yet been repaid for the credit issued.
it's because with credit cards they can check the credit limit, then be sure that the card can pay the insurance deductible in case of crash
instead with debit i can rent a car, close or deativate the card, crash/total the rental car and then avoid paying any extra fee
most rentals don't rent with debit cards because they want to be sure, and who accepts debit:
- they preauthorize thousands of dollars instead of hundreds
- they only rent the lowest end of the available cars
-
Oh, so Hertz has gotten wise to... every online platform that exists: Outsourcing all responsibility for their user-hostile bullshit to some vague "system" that cannot be held accountable.
I'm so sorry but the advertised cost has doubled because... Computer says so! No, sir, there's nothing I can do, sir, you see it's the system.
And you can't go anywhere else, because everyone else is doing it (or soon will be) too!
Oh, this is a thing. It's called an accountability sink.
There is a really interesting book called the unaccountability machine by Rory Sutherland (if my memory is working). Worth a read
-
And what is processing that information?
Computer vision commonly uses convolutional neural networks on the input, which is different from the transformer neural networks used in LLMs. If you have more info indicating LLMs are used here please share
-
in this case, hertz doesn't rent to who doesn't have a credit card
debit = no rent
I should have remembered that. I had to lend my card out to my friend who was in a credit lock at the time they needed a rental. Still, I don't think my advice is invalid, just irrelevant here.
-
Oh, this is a thing. It's called an accountability sink.
There is a really interesting book called the unaccountability machine by Rory Sutherland (if my memory is working). Worth a read
Thanks! I knew there was a term, but just couldn't conjure it.
-
You are spot on here. AI is great for sensitivity (noticing potential issues), but terrible for specivity (giving many false positives).
The issue is how AI is used, not the AI itself. They don't have a human in the checking process. They should use AI scanner to check the car. If it's fine, then you have saved the employee from manually checking, which is a time-consuming process and prone to error.
If the AI spots something, then get an employee to look at the issues highlighted. If it's just a water drop or other false positive, then it should be a one click 'ignore', and the customer goes on their way without charge. If it is genuine, then show the evidence to the customer and discuss charges in person. Company still saves time over a manual check and has much improved accuracy and evidence collection.
They are being greedy by trying to eliminate the employee altogether. This probably doesn't actually save any money, if anything it costs more in dealing with complaints, not to mention the loss of sales due to building a poor image.
If it's fine, then you have saved the employee from manually checking
Exactly. Not only that but the human is more likely to overlook some things. It also creates a digital record of the complete condition.
Have the AI go over the vehicle, being insanely meticulous and then pass that info off to a human who verifies any flagged damages in a couple of seconds and makes decisions about what needs to be charged.
Combining the 2 improves efficiency and accuracy.
-
Oh, so Hertz has gotten wise to... every online platform that exists: Outsourcing all responsibility for their user-hostile bullshit to some vague "system" that cannot be held accountable.
I'm so sorry but the advertised cost has doubled because... Computer says so! No, sir, there's nothing I can do, sir, you see it's the system.
And you can't go anywhere else, because everyone else is doing it (or soon will be) too!
I once tried to book on their site, and the website froze and it wouldn't go through every time I tried to pay. I checked my email multiple times and checked my credit card statement. Nothing went through. I went and booked elsewhere. 12 hours later, I get a confirmation email from them. I tried to cancel and it wanted to charge me $100 cancelation fee. I had to call to get it resolved. 45 minute wait time. Thankfully they took care of it, but it was a huge headache caused by their shitty system.
-
This post did not contain any content.
"AI is a disaster." Fixed it for you.
-
Computer vision commonly uses convolutional neural networks on the input, which is different from the transformer neural networks used in LLMs. If you have more info indicating LLMs are used here please share
If you have more info indicating LLMs are used here please share
two seconds of research would reveal LLMs are ALL OVER COMPUTER VISION. Are convolutional networks used? Yes. Are LLMs used? Yes. And MLLMs.
Tell you what sparky: you find me a source that says ONLY CNNs are used, then you can act like a subject matter expert.
Large Language Models Meet Computer Vision: A Brief Survey
Abstract page for arXiv paper 2311.16673: Large Language Models Meet Computer Vision: A Brief Survey
arXiv.org (arxiv.org)
It's not just words: LLMs in computer vision | Microsoft Community Hub
Discover how LLMS enhanced Azure AI Vision service and unlocked new scenarios, like image generation.
TECHCOMMUNITY.MICROSOFT.COM (techcommunity.microsoft.com)
GitHub - OpenGVLab/VisionLLM: VisionLLM Series
VisionLLM Series. Contribute to OpenGVLab/VisionLLM development by creating an account on GitHub.
GitHub (github.com)
https://www.chooch.com/blog/how-to-integrate-large-language-models-with-computer-vision/
-
Oh yay another bullshit thing. No thanks.
It's been around for quite awhile. I use Turo more than I use regular car rental services because you actually get to choose what car you're getting and the prices are better.
-
What's different is that most people will see it as "tech stuff" and mentally file it in a drawer with spare extension cords and adapters. They don't care to deeply study or catalog things. Nerds care about that, and most people here, including me, are nerds, but most people are not nerds and consider learning to be a form of torture.
People writ-large don't care about proper genre labels either, they just kinda pick a vibe and guess off of it. Look at all the -core suffixed aesthetic names that cropped up in the last decade.
Yeah, I think it's unfortunate that tech is something people refuse to learn about. I've been able to explain technical topics to less technical people, they just need to care.
For example, I'm into finance, and have been able to explain pretty complex topics (compounding, Social Security benefits, derivatives, etc) to people with no background in a way that they know how things work at a high level. They may not be able to trade options or predict portfolio performance, but they can at least tell if their "financial advisor" knows their stuff.
Learning a bit about key technologies can help cut through the BS from marketing departments. But as soon as I mention something remotely technical, people shut down. If people understood that LLMs basically do keyword association to generate text from a prompt, they wouldn't believe the lies that claim they "think." Just a little bit of high level knowledge would change it from "magic" to a sometimes useful everyday tool.
-
Oh, this is a thing. It's called an accountability sink.
There is a really interesting book called the unaccountability machine by Rory Sutherland (if my memory is working). Worth a read
Rory "expert in all things" Sutherland?
He keeps cropping up in my youtube feed talking about a huge range of topics in his confident posh twat voice.
His background is in marketing... never trust a salesman
-
If you have more info indicating LLMs are used here please share
two seconds of research would reveal LLMs are ALL OVER COMPUTER VISION. Are convolutional networks used? Yes. Are LLMs used? Yes. And MLLMs.
Tell you what sparky: you find me a source that says ONLY CNNs are used, then you can act like a subject matter expert.
Large Language Models Meet Computer Vision: A Brief Survey
Abstract page for arXiv paper 2311.16673: Large Language Models Meet Computer Vision: A Brief Survey
arXiv.org (arxiv.org)
It's not just words: LLMs in computer vision | Microsoft Community Hub
Discover how LLMS enhanced Azure AI Vision service and unlocked new scenarios, like image generation.
TECHCOMMUNITY.MICROSOFT.COM (techcommunity.microsoft.com)
GitHub - OpenGVLab/VisionLLM: VisionLLM Series
VisionLLM Series. Contribute to OpenGVLab/VisionLLM development by creating an account on GitHub.
GitHub (github.com)
https://www.chooch.com/blog/how-to-integrate-large-language-models-with-computer-vision/
I was actually referring to UVEye which was referenced in the article. I looked into UVEye and nowhere did it say it used LLMs with their computer vision. That’s why I asked if anyone had any info on them using it. The comment I replied to assumed LLMs were used but supplied no evidence. None of the links you shared have anything to do with UVEye either.
-
I was actually referring to UVEye which was referenced in the article. I looked into UVEye and nowhere did it say it used LLMs with their computer vision. That’s why I asked if anyone had any info on them using it. The comment I replied to assumed LLMs were used but supplied no evidence. None of the links you shared have anything to do with UVEye either.
Computer vision commonly uses convolutional neural networks on the input,
no where do you specify UVEye.
You could admit they're all over, but instead double down on how I assumed lol
-
Computer vision commonly uses convolutional neural networks on the input,
no where do you specify UVEye.
You could admit they're all over, but instead double down on how I assumed lol
Except they are using computer vision, not an LLM
That’s what I initially said, referring to the article. If you have nothing to say regarding the technology in this article that’s fine, but don’t just assume that since there is research of incorporating LLMs into computer vision means it was used in this specific case.
