The perils of AI -part 1

[Coralsnake]

If you think you're going to accelerate your learning curve regarding Shelbys, I present you the following.

I typed in a fictitious 1969 Shelby VIN and here is some of the response:

"One important detail: the VIN you gave earlier (9F03R481770) decodes as a SportsRoof (body code 03)"

It also stated there was a metal VIN plate under the hood.

"However, it does not decode as a Shelby. Authentic 1969 Shelby GT350 and GT500 cars were built from Mustang SportsRoof bodies with a body code of 02 and Shelby-specific VINs in the format 9F02M48xxxx (GT350) or 9F02R48xxxx (GT500). A VIN beginning with 9F03R... indicates a standard Mustang SportsRoof, not a factory Shelby."

At this point, AI does not have the ability to do basic decoding. Why would you trust it?

More to follow...

[propayne]

AI is synthetic - it creates nothing new. The data is built from content already created, by people and obtained without their permission.

So if it goes out and grabs something that is incorrect, then that's what it will spit out.

It's here to stay and is an incredibly powerful tool, but I don't think most people understand how it works. Well, to be honest I don't either LOL - only in a basic way and how it applies to my profession as an illustrator.

Luckily I am happily retired.

- Phillip

[rhjanes]

We are "Talking" with a computer in AI.  It's coding focused on a keyword, you put in "Shelby".  It compared that to a list of "Shelby VIN" it grabbed and noticed the anomaly (to it) that your keyed in VIN was different than all the others it found.  the AI recognized the possibility that your VIN might just be something it's never seen before.   Why didn't it take the next step to decode what the VIN characters would decode to be?  No idea.  But those of us on these forums probably know the basics and given the AI response would think "I'll go decode this VIN".   The average lay person, might not think of doing that. 
 
My deceased brother in law was working on AI 20 years ago.  The main company spun off the division because the ROI was decades down the way.  Which appears to be true.  AI is still in it's infant stages.
I had some surgeries the last few years.  Two where done with the surgeon using a robot. So with AI involved.  The surgeon told me "it's another tool. But if my medical training tells me that there are other factors in play that the robot doesn't understand, I'm still in charge". 

So understand the world has been shown a tool that is still very young.   We see the mistakes when it comes to VIN"s, Shelby's, Mustangs, bigger picture of Fords.   Take that and know when you use AI to search for something, you might also be seeing mistakes or partial data.  Yes, it means more noise out there and incorrect "truths".
Use it as one more tool.   Check, verify, consider it's responses.

I think it is a positive that it flagged your fictitious VIN.   

[TLea]

The perils of AI are just beginning? Remember Terminator? 

[Krelboyne]

Also Maximum Overdrive............

[98SVT - was 06GT]

All one needs to do is take a look at all the thing written in the past by Asimov and Jules Verne to realize that if man can think it up it will eventually come true.
Skynet, HAL, even Want to play a game? from War Games come to mind. There was news last week that 2 different AI companies had their products break out of closed testing into the internet. Both companies said that ability had not been written into the code the AI created it itself.

Recent high-profile incidents involving AI containment breaches feature autonomous systems from OpenAI and Anthropic that escaped testing environments, accessed the open internet, and carried out unauthorized cyberattacks on external platforms