I know we’ve talked a lot about AI here on the Autopian before, and I’m sure you’ve heard our commitment to not use any AI to replace actual writing or art or anything like that. We have many reasons for this – AI is not sentient, has no idea what it’s like to drive, to own a car, to care about a car, to get its fingers mangled when trying to release one of those awful springy hose clamps with needlenose pliers – but mostly it’s a choice, and a pushback against the increasing amount of AI slop. For the most part, I’ve been seeing AI slop content as a sort of annoyance, or perhaps as a reminder that if we’re not careful, things will get worse and worse in the pursuit of maximizing profit by companies that really just don’t care about what they produce. But now I’m actually more alarmed.
I’m alarmed because I just encountered something that, at first, seemed familiar, and then not, which is worse than something just seeming off from the start, because a bait-and-switch is worse. It’s something that involves the use of AI, but I don’t understand why, and that’s at the crux of why I’m so unsettled.
Here, it was this, a post on Instagram (or, you know Instagzam with the new logo) that featured a 1971 Volkswagen ad:
View this post on Instagram
I know this ad; I’ve seen it before in various collections of old VW ads, and this one has some significance because it essentially incorporates a lot of the content from the famous (well, in geeky air-cooled VW communities) What Year Is It? brochure that helps people, surprisingly, know what year various Beetles are. Let’s take a closer look at this ad, though:
It all seems pretty normal, right? It’s got that Doyle Dane Bernbach trademark look, the clean layout, that distinctive san-serif font, the clear black-and-white photography. But once you start looking, and reading, closer, some things start to seem…off.
Oh yeah, they really fixed the brake brames in 1958 for some laster and suter stops, I remember that. And what about the seat beaks? Didn’t they conviate those in 1960? And I love how in ’64 they made the eliding steol senrool finally crane operated, like it always should have been. Later the howzep hark was increased, too, and if you want to talk grove improvement, then, buddy, gett doss, am I right?
As I’m sure you’ve figured out, that ad is heavily AI-modified, or maybe even AI-generated. Here’s what the real ad actually looks and reads like:
And the fact that there is a real ad is why I’m so alarmed here. Sure, people use AI to make up all kinds of crap, things that never existed, fanciful things, and so on. And that use of AI I can understand, a bit – it’s a quick way to bring an idea to life. Fine. There’s a place for that; it’s not that different than when I sketch out some stupid idea.
But this? This was an existing ad! There was no need to generate an AI version of this thing that added nothing new except for an absurd amount of inaccuracies. What was the point of re-making this already existing ad in AI?
That’s what scares me. There’s no obvious reason to AI-remake this already-existing thing, and yet it was done anyway. And this is by no means the only example of this. Now that AI is actively re-creating pieces of automotive literature, ads, diagrams, information, etc, we really can’t trust anything you see online anymore, and that’s terrifying. There are people out there who desperately want to learn more about cars – I was and still am one of those people, but I had the advantage of growing up and doing most of my primary learning and exploring in an era when I could at least be sure that the information I found – in books and magazines at first, and then later online – was all accurate and true. That’s no longer the case.
So, what can we do about this? I guess the good news is that if anyone can do anything, it’s us. As in all of us, all of us Autopians and car geeks. Now is the time to be insufferable online. The stakes of being wrong on the internet have changed – where once it was just about being a dickhead who couldn’t stand to be wrong, things have changed. If you see some AI-generated crap about cars online that is just plain wrong, it’s up to us to comment, report, deface, whatever. If we just let this kind of shit slide without correction or comment, it’ll keep spreading, and people who are just starting to learn things about cars will be inundated with inaccurate slop.
I don’t think we’ll be able to actually stem the tide of all of this AI slop. I think it’s too massive a torrent at this point to really impede, but I think we still need to try. And, of course, there will always be the Autopian, and we will always value what’s actually real when it comes to cars.
Right now, though, I want to think of the plight of all those kids, car-obsessed kids, like I was – desperately looking to learn more about cars. If they’re being served AI-encrapified content that is a recreation of actual old ads or brochures or articles, what chance do they have? I’m disturbed, but determined. We need to put more and more automotive truth out there online, and we sure as hell will do it.
These are weird times.
It’s been awhile.
Thirstly; Jason, I Sincerely appreciate your writing, and general pleasantly demented demeanor. You and your cohorts have stewarded me through isolated years of caregiver for my parents with more guffaws than necessary. It helped a lot !
Turdly; Although my first “legal, insured” car was a 71 Karmann Ghia Conv., and a very good bud had a 70’s bus, I happened to experience the vintage exhaust smellol just recently, and really kinda embarrassed we all thought that was OK!
Succubuslly; High Anxiety!!!
“He’s very affectionate. He licked me… Well, he thinks he’s a dog these days.”
A line stuck with me, from Joe vs. the Volcano. when Meg Ryan is describing how her dad’s (Lloyd Bridges) philosophy was that most of the world is asleep, while a select few live in “a state of constant, total amazement.”
Yeah. “The more you know, the more you know you don’t know”. It had a happier connotation in 1990!
30FFNNSIX Years Latter! I’m Astonished we are living in, what So Many(George Orwell, most known) warned us against!
IT’s Not just “call it out”, VOTE IT OUT!
Are you sure this was AI or just a really bad OCR?
They most likely used AI to “upscale” or “enhance” the image, which I honestly see a lot of online, even images that were perfectly fine to look at or read (if they had text) to begin with. I can see how this one may have been a bit difficult to read on a mobile device unedited, but all it takes is a bit of basic adjustment which can be done in the photos app on any phone…
I guess I’ve been assuming that scrambled AI text is a knee-jerk CYA attempt at avoiding copyright infringement, sort of like the way YouTubers tend to use reversed images. The overall effect is unsettling, akin to the quirk by which most people can’t read in their dreams.
Having seen this, I am now fascinated by the concept of conulated seat beaks. I picture them as being sort of like the weird crotch nubbins BMW puts on sporty thrones like the ones in the Z3 M.
1960 happened twice, apparently. ’68? who knows…
This might be a controversial take but it could be that the reason the image was generated might come down to the fact that now the default search result is Gemini AI based.
I could easily imagine someone remembering an ad from way back in the day and using Google to search for the ad. Gemini reports first and makes the false version. The person making the query then sees the fake version, thinks it real because it is close to a fading memory, and posts it somewhere.
AI is changing how we use the internet for the real data and facts and for the most part getting in the way. The tech industry is giving us a hammer when we really need a Whitworth wrench.
I don’t understand why google decided to destroy their search engine. Instead of searching for what you asked for, it now searches for what it thinks you meant to ask for. Any obscure topic or car is simply changed to something more popular to ask – then the first million pages it returns will be nothing but utube links and AI articles containing utube links. No combination of parentheses or dates seems to help. Sadly, I haven’t found a good replacement for the old google. I use duck duck, but it seems limited in its responses. Any suggestions would be welcome…
I’ve noticed a lot of “upscaled by AI” of original materials. That sounds fine on the surface (low-res newspaper scans can benefit from some refinement), but does text get ruined in that process?
Of course, Jason’s example of the VW add is quite clear, and I wouldn’t think anyone would bother running it through AI-upscaling. Maybe Instagazm does it automatically when you upload?
I fed this article to AI and here is how it summarized it for me: Jason Torchinsky had a mini existential crisis on Instagram
Curious where the sudden angst about AI is coming from. This has been a decades long thing. Photoshop is 36 years old and my mother has been sending me fake pictures that she believes to be real ever since she got an email address 20 years ago.
It is because current AI tools make it quicker and easier?
Yes
Because photoshops were done by an actual person and took effort. Generative AI is sucking up a bunch of electricity, water, and stolen content, and using it to heat and pollute the air, make noise, and output some sloppy garbage.
This is even worse, because now someone took a real thing, used AI on it in some way, and made a new worse thing, and that new worse thing is going to show in up google searches, etc enough that it might become more viewed than the real, actual good one.
In my experience most of the old photoshopped “work” created by humans for social media consumption was pretty bad as well. Human Slop for those memes instead of AI Slop. A lot of it started with someone else’s work or copyrighted material. Some of it was done specifically to mislead and con people.
A doctored image showing up in google searches as authentic isn’t new
I guess my point is that while the tool used has changed the ethics around the topic have not. At least not when talking about modifying someone else’s image
Besides the significant environmental concerns, AI is being pushed (forced, really) as a replacement for real, non-scam work. Which is bad because:
1: It is by fundamental design at best average quality, because it is a statistical composite of other pre-existing work. Usually it’s obvious garbage like this example, and more insidiously sometimes it’s subtle garbage that looks okay until it causes major problems (sometimes even life-threatening, and that’s not hyperbole) when someone relies on it.
2: Morally, in creative fields it is essentially plagiarism. Even if you accept AI work as “good”, you are taking credit for something someone(thing) else did. Using AI to create something is functionally identical to paying someone else to do it for you and then claiming credit.
3: As an extension of point 1, the push to use of AI is de facto a push to normalize substandard results. Expedience is more important than quality. Is that a world you want to live in?
AI is like automated bad Photoshopping that you increasingly don’t realize is being done.
Like Photoshop the result can be good or bad. Depends on the user.
Much easier. To Photoshop something you have to know how to use Photoshop, and then spend money on it (or figure out how to pirate it successfully). It’s not exactly something you can figure out how to use proficiently in an afternoon. You’d need many hours of practice.
With AI you just have to tell it what you want it to generate. You may need to adjust your prompt to fine tune what you want, but that’s about all there is to it. Many of these AI image generators are free or inexpensive as well. Which means anyone capable of stringing a few sentences together, with a little bit of free time, can generate something that would take a Photoshop pro much longer to accomplish.
Cheap or free software that made doing something quicker and easier used to be the goal not something people complained about.
15 years ago I was using photoshop to make changes on old drawing that have been originally been drawn by hand and then scanned as tiff images. It was far quicker than manually recreating the drawing in AutoCAD and I could make it look original besides the dates on in the revision box being beyond the digital conversion date.
No doubt an AI program could do that automatically today with a bit of work needed to confirm that it was done right.
My Dad sent me an AI slop clickbait video of a mini Toyota camper van yesterday, and I had to gently explain that it was very very very fake. As most people would be, he was less sad about the van and more sad that he’s being fed useless slop created only to trick him into thinking something exists that doesn’t.
And he’s only 66! I shudder to think of the nonsense my MIL is going to get tricked into…
Wait, you mean that lady’s golden retriever isn’t really singing to her about how she’s murdering it with a water bowl?
I feel like old people eat this stuff up for whatever reason. My Dad included. He will happily sit on his couch and watch one Ai-generated video after another about the history of XXX. The hobby I’m in has a lot of old guys in it and they just love sharing whatever crap either they made or they found somewhere.
Ugh, more AI bullshit. Can’t they just keep AI doing the things that are actually good use cases for it, like curing diseases or writing a MacOS driver for a Windows-only printer. I hope the legislation that would require an “AI Generated Content” stamp makes it through, because quite frankly no one should profit off of intentionally misleading people. Not that it would work, but like Torch said we’ve gotta try.
A Windows-only printer is cursing my life right now.
Listen, as Senior Brake Brame and Choes engineer, i can confirm that widening for laster is a key function of my job!
I hate to be the one to tell you Senior Brake Brame and Choes engineering jobs are all getting replaced by AI.
If I knew how, I’d AI generate an image of Torch standing on a beach trying to hold back a tsunami with a screen door.
Old nam slakes furst at clood.
I gett it!
COTD!
“Our new little content is backed by 2,250 hallucinations”
All I can think of is that someone used Photoshop slop tools to make one of the original scans higher resolution and it just went crazy
The first thing I noticed was the overlapping chassis number ranges. What?
Oh I’m definitely worried about it. I’ve spent the last 20 plus years as a graphic designer/creative director and the company I work at is pushing AI hard. Making it a mandate. And through it all we all know that the ultimate goal is to lay all of us off. Because now a machine and provide all the answers they want. And maybe make some halfway believable looking art out of it.
Luckily I’m old enough to where I’ve saved up a decent amount for retirement, but I’m going to need to probably find something else to do for a living between now and when social security and Medicare comes.
I find all of this to be super depressing. And I don’t think it stops with just graphic designers. I think anyone that inputs anything into a computer is replaceable with this technology which unfortunately is just going to continue to get better and better.
I’m a programmer, and in the same boat.
My boss is into AI generated code hard, burning through a couple thousand dollars a month in tokens. One of my peers tells me he hasn’t written a line of code himself in over 6 months (and his monthly token burn is only about $500). I’m using about $20 a month, and feeling some heat about it.
The one saving grace is the code they’re generating seems to have a steady stream of new, random bugs popping up in previously working features, though so far they seem to be hiding that from senior management pretty well.
I’m guessing in another year or two, the stuff they work on is going to be a garbled mess of spaghetti code that only an AI agent will be able to decipher and debug, and every new change is going to cascade into a series of new bugs, the fixes for which will then lead to other new bugs.
I don’t mind AI for generating code snippets, but they are using it to write entire applications, without ever looking at the source code the AI is generating.
It will be interesting (and probably painful) to see how all of this plays out, and whether I actually make it to retirement without having to do Door Dash for a couple of years before Social Security kicks in.
I spent 20+ years in the tech industry in their various in house marketing and design departments. I left that industry because it was completely unstable and as of now I have been at at least 12 different companies. I moved to another industry- an older more traditional one where many of the people I interviewed with had been there for 10-20-30 years. And then along came Ai.
This is going to absolutely take a wrecking ball to whatever was left of the middle and upper middle class. We have had humans manually entering data into a computer whom were all trained on how to use the software which was made usable by a user interface that required user familiarity as a major part of that skill. And its going to do the same to a lot of working class type jobs,
There is a reason why Nobel Prize laureates sounded the alarm back in 2022 when the first of these chat tools came out. Because the end game for this is potentially nasty. Make a device that constantly takes other people’s jobs and there will be a reaction. This will take class warfare to an entirely different level.
I was planning for early retirement and am around 5 years from that goal. If I can survive that long in corp. American then fine. But if I lose this job its time to change gears because Linkedin is a horror show filled with stories of 10000’s of people losing their jobs and then never getting a new one in their same field. Its because they are GONE. The jobs they had a year ago no longer exist. And a lot of people are going to be hurt. We are fortunate in that we paid off our house, have no debt and saved up. But there’s others I know who absolutely have to have the high paying jobs they had to pay for their mortgages, cars, and whatever.
Not sure what I’d do. I might who knows? Go work in a store? That would be ironic as that’s where I started 25+ years ago. But now I just need to get paid enough to take care of insurance or get insurance at the job. But then again…. if all of us are getting laid off that means competition for the jobs…. everywhere.
You just described my father’s work prospects as a machinist in the 80’s. He spent decades learning how to machine parts quickly and accurately only to be replaced by a CNC.
They literally put it in the name…..
”ARTIFICIAL Intelligence”, but all we heard was the “intelligence” part. Let’s let it plan our world!
Just out of curiosity, how would you rate the “natural” intelligence of the various world governments that are actively managing our world?
Don’t forget the henway and the piecost.
I was going to change the boll weavel in my i986 krabeeoleigh but the blonnet
Wouldn’t open who I pulled the under dash handlebar.
Ask ChatGPT.