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.
I have stepped off Instagzam, to not be exposed to, or generate views to AI BS.
Trying to find a way to avoid FB too, but many of our old car social activities are organized through that 🙁
Have my “curated” (by me…) subscriptions on YouTube, so not going to quit that, but I love clicking “don’t recommend channel” to punish general stupidity and the algorithm (all hail 😉 )
About a year ago I canceled my Facebook, Twitter, and Reddit accounts. I never had Instagram or anything else really. And even though I used to visit Facebook around once a day? I haven’t missed it or any of those since I quit and I think a lot of people would benefit if they did so too. These social media companies are really not worth your time anymore
It is pretty nice that you can turn off your watch history on YouTube, and it’ll totally turn off any “feed,” allowing you to just look at your subscriptions
Yup. I got that rec from Technology Connections a while back and it’s been great. Don’t let an algorithm decide your interests for you.
So can we start calling it Artificial Stupidity now?
More like Artificial Ignorance
This sloppy application of gen AI is poor. The risks of gen AI are very real, and we don’t even understand what all the risks are.
All that being said, in the right hands with the right training and supervision, some gen AI tools can legitimately improve certain fields (if you ignore the environmental costs and other “externalities” much like the world didn’t do much about the increased emissions and pollution when the industrial revolution happened). I’m a software engineer, and gen AI helps me in at least two ways, with at least two important conditions.
1. A lot of code is boilerplate crap I’ve written hundreds of time. Like “connect to this database, pull data, write it to a file” type code. Gen AI can do this faster and I can focus my brain on the more useful stuff. I know it’s easy to make fun of software engineers for having a pretty cushy work life, but using your brain to create software is taxing and there’s only so much you can do well in a day. So if the machine can do the busy work, I can spend more of my brain on the useful, creative stuff.
2. Gen AI tools for programming are spookily good at creating automated test suites. They reason about the code, figure out where it can fail, and write tests to ensure everything works, every time you make a change. Humans suck at this. It’s hard to reason about failure points, and writing tests is boring as hell, so most engineers take shortcuts and don’t write enough (or any!) tests. In that way, gen AI has increased the quality and reliability of the software I write.
The important conditions:
1. Whatever I produce, whether I wrote it or the machine did, is my responsibility. I must read it all, understand it, and stand behind it. I’m not pushing AI-generated code without my imprimatur. If it breaks, it’s on me.
2. I instruct the tool to tell me what it plans to do before it does it, and grant permissions step by step, so I know what’s happening as it’s happening (assuming an agent isn’t fking around in the background sending my memes to the Chinese government). I frequently tell it to do things differently. This goes back to point 1.
Used in that way, a gem AI tool is a lot like a CNC machine or a 3d printer or a loom. The quality depends on the operator’s skill and the tool’s quality. The results are quick, predictable, and repeatable.
But most importantly: the human comes up with what to do. The machine does the grunt work. As long as the human initiates and guides the machine’s work, and again ignoring the externalities for the sake of discussion, it’d be foolish to just say “AI bad” because that’s objectively not true in terms of its output (code, medical applications, you name it).
Now if you bring the externalities into the discussion (power and water use, noise from data centers, pressure on component prices that hurt consumers, etc), then I don’t have any objections to people having a hard-no stance against gen AI and refusing to use it altogether (if possible).
Ignoring the externalities shouldn’t be encapsulated is some parenthesis and set aside. That’s like saying Mussolini made the trains run on time, asbestos insulates great, and lead really makes the paint colors pop!
You can’t try to take the good (I produced a simple procedure in 10 seconds instead of 10 minutes) without the bad (massive increases in droughts, collapsing glaciers, and gross income inequality). It doesn’t work that way.
You need to take your own advice and adopt a more rigorous stance until the externalities are resolved.
This should have been COTD
I used to edit newspaper articles rather than code, so I can relate to this perspective.
“Churn, but verify” is an approach that makes sense.
> What was the point of re-making this already existing ad in AI?
Enhancing the car images when you don’t know how to use Photoshop + prompting poorly + not checking the output.
At least one of these is laziness or pressure from a herb to crank stuff out faster. At least two are lack of training.
So, a federal court said that AI companies can slurp up books and it is not a violation of the copyright. However, my understanding is that material generated by AI cannot be copyrighted. So, if AI slurps up a copyrighted ad and generates something, is the output copyrighted?
Asking for a friend.
I had this experience a while back when I saw a very clearly AI generated version of this image somewhere online. Unfortunately I can’t seem to find it now, but it had the typical hallmarks – the store names in the background were not in real letters, the cars in the background looked vaguely familiar but not, and the car in the foreground had changes to make it look “off.” It took me a minute or so to think “Wait, haven’t I seen a non-AI version of this somewhere and of course I had. That’s when I had the same thought. Why? What is the point?
AI has been a hot topic for me lately. I am a very reluctant adopter but not completely a luddite whereas my wife works on implementing AI for her company. We’ve had a few spirited discussions about this, and now I know how the Conways feel. Also, this week was my first time dipping my toe into some AI tools provided by my company and it was an abject failure. I provided negative but objective feedback to the team that developed it and one of them said that my feedback validated several concerns he had. Vindication I guess?
Woah, wait! What’s that photo’s backstory? Why the hell are they clobbering a 3rd gen Corolla exactly like mine?
Why did you show me this? I’m not crying, you are crying!
“In 1982 in northern Indiana, two local businessmen took part in an unusual event: they smashed a Japanese-made car. In this economically hard-hit steel region, union leaders organized a fundraiser where people paid $1 for each swing of a sledgehammer at the car. The proceeds helped support families.”
Oh man, it’s worse than I thought. Thanks for the context.
Because instead of blaming domestic automakers for treating their customers like garbage, thus driving them to buy cars built by companies who actually cared about building quality products, it was easier to just blame the competition and take anger out on that
I’ve been seeing a lot of older (or even a few years old) automotive ads and pictures being “upscaled” by AI for the modern internet. It also happens to completely ruin them in the process.
AI AIn’t real.
This looks like someone had a low resolution version of the ad and had an AI upscale it. The text was probably too pixelated in the version they had and the AI guessed at a lot of words.
I tried to do this with some blue prints of the Haunted Mansion at Disneyland. It did successfully upres them but the text was jibberish like this.
I sure as hell didn’t put the result on the Internet though.
Zoom and Enhance!
Howspower!
That’s a prime example of Zuckerberg turning up the slop faucet on Instagram. He’s trying to drive up demand for AI apps, but demand remains pathetic. And meanwhile their user base is shrinking, not growing.
What’s worse for him is he’s investing hundreds of billions in AI infrastructure, and it’s driven away investors. He’s killed $500 billion in market cap over the past year.
https://www.businessinsider.com/meta-alphabet-stock-price-decline-record-highs-mag7-nvda-earnings-2026-8
The core business of Facebook is solid due to their monopoly. Ads and Marketplace remain healthy. But it’s obvious to everyone but Zuckerberg and his toadies that AI is a disaster and the sooner he trashes it, the sooner he can recover.
So the right answer is to get mad at AI, or laugh at it, but don’t panic.
I’m getting a performance review tomorrow that’s largely negative, focused on my not using AI enough at work. It is not an exaggeration to say I fear being fired. As much as I hate AI, and as stupid as this whole bubble is, it’s not all fun and games, unfortunately.
The guy is a dumbass. And I’m amazed that investors keep throwing in money because he literally pissed away billions of dollars on that meta project, literally changing the name of a company over a failed project
Challenge:
Intentionally generate an AI slop Warm Start to see if we notice.
I wrote a book once ” A History of the London Lead Company” , it sold in dozens.
Lots of tables about yields and costs, a very very boring book.
I have a train set that is an idea of an idea of what might have happened if Lead mining was still a thing in the 1980’s. The model railway is a joke, a pretty toy but not real although I wrote about building it as if it were.
I also wrote a book about Victorian attitudes to death, that sold better than the others. Oddly enough the charnel house at Leadgate was not part of the Necropolis Railway, it was (and most of it still stands) a storage facility for corpses awaiting burial in consecrated ground. It is in the high Pennines, near Alston. To reach a goodly planting spot it was necessary to traverse fiends fell. Those who walk the Pennine Way will know it as Cross fell.
AI really fucks this up.
Dozens! Impressive given the subject matter. And important work, because without people like you, entire chunks of knowledge would disappear forever. When a culture forgets something, it can never come back.
Thank you.
You can’t even search basic car information on Google, the first 18 articles are all Ai slop with 15 paragraphs of information when you just needed the oil capacity for a delta 88.
All. My car searches I have to add “forum” in the search to get real people’s answers.
I cannot stand the amount of AI generated slop I have to see on a daily basis.
Except for the Lego vids coming from Iran.
Those are GOOD.
It breaks my heart when my high school history teacher – who was freakin’ brilliant, and who I wanted to keep in touch with long after I graduated – sends me yet another Facebook Message with some AI slop kitten or political image.
Looks less like AI and more like someone converting a PDF to Word.
It’s called Insta-brag. Not Insta-gzam.
Yet another site I stay far, far away from like You-Tube and Farcebook.
Filled with influenzas. Some people call them influencers.
That big “23” not being Futura (well, that, and the superfluity of the big “23” in the first place) should have been all the tip-off needed to know this wasn’t a DDB original.
Like all those “rowdy statement” fake Porsche “ads” where they didn’t even use Franklin Gothic, like any normal person would.
The best lies are those close enough to the truth that you question yourself. This zombified ad is probably a result of upscaling gone wrong, and we could spot the slop this time – but soon those algorithms are going to proofread their slop enough that they will check for coherence, then something like this will pass with “upgraded grille” instead of “upgrade brakes” (because AI will probably never check if a VW even has a grille).
AI content is like boxed cakes, fine if you are home cooking in a pinch, not fine if you are doing it professionally as a career. I am glad we’re served only the freshest cakes made with love from rust flour, perfectly aged engine oil and only the finest transmission shavings.
I hate this timeline, can I switch?
Sure, we currently have two other timeliness on offer.
1) starving peasant in the 500’s and
2) fish merchant in Constantinople in 1451
Orphaned Gunter in 2045?
Oh wait, we’re trying to avoid that timeline.
My guess is, they had a low-quality copy and used AI to “upscale” it. It did fine cleaning up the big stuff, but the small letters were blurry/patchy and so it invented garbage to fill in what it couldn’t “see” clearly.
Who knew that an old Encyclopedia Brittanica may someday be worth more than gold. I still have hard copies of my old engineering school texts. I may be rich!
First edition machinests handbook has more information in less space then later editions. Anything past the 4th is just textbook nonsense.
I have one… have to check the edition. At least old textbooks are not AI slop and basic material science, mechanical and electrical engineering principles have not changed.
I have a bunch of old ’50s vintage techincian level training manuals on everything from hydraulics to HVAC. I think I have a manual for an old slide rule as well (never, ever had to use one as I’m Gen-X).
I keep a collection of old technical books as well. Blasters handbook also very interesting. Along with the boiler operator handbook. Sometimes old guys will give you things you feel like you aren’t supposed to have. I’ve been given some interesting technical data and manuals from different industries.
I met the text book Mafia about 15 years ago. When they pushed my friend who was a dean into a dinner. What a group of ignorant people pushing people to buy their slop. I’m not sure if AI will make it much worse but probably. Still hasn’t been good in decades.
I was given a slide rule by an old friend he said some of the big time nerd guys used to have special pouches for them they would walk around with. Must have been amusing to see. I just remember some of the guys with the hp calculators in their shirt pocket.
My dad had a special pouch / slipcase for his slide rule. He was originally going to be an architect and I guess he liked the rule.
I’m hanging onto my red white and blue Funk and Wagnalls!
Not only did I keep all of my engineering books, I’ve continued to expand my library. I now have two shelves at work made up of books I’ve purchased along with binders full of reference materials I’ve compiled on my own over the years. The company’s newest hires don’t seem to have much in the way of books. 🙁
I’m much the same… much of the office library is mine.
Screams open AI, but I bet it was the meta ai that is even more hot garbage. Feedi seat backs seems like a much more modern opinion then 56.
This story gives me ideas as to how the current administration submitted a map where they got every single country wrong in Africa
At least we can rest assured that our nation’s capitol remains securely in North Carolina, well south of Lake America.
https://www.the-independent.com/news/world/americas/us-politics/trump-lake-ontario-america-name-b3039587.html
Did it also connect the muffler bearing to the rezuberator?
Don’t forget to top off the turn signal fluid.
Hey it’s not a BMW!
Even though it’s listed for 1969, my ’74 and ’77 still didn’t have a day/ night reainixe rooror so I have to assume some of these changes were rolled out at different times for different markets.
encherining
No, they were both regular sedans.
My best guess is someone ran it through AI to increase the contrast. The original one looks dull and faded in comparison. However the original poster is also still eminently readable, so yes, this was completely unnecessary.
A lot of people I know sort of assume that AI will just inherently make things better, so they run a lot of unnecessary stuff through AI – pictures they take, literally everything they write/post, etc., and don’t bother to double-check any of it before posting it.
This is likely. I’ve only used AI once to create an image for a birthday card for a friend. What disturbed me is AI chose to make modifications to the image that I did ask for nor want. I’ve never returned to using AI for imaging. I try not to use it at all, but it is difficult to avoid entirely.
Same here… while updating a fireplace mantel in a room, I would give ChatGPT a picture of the room, and then tell it, “Show me the picture I just gave you.” and it would move a door just a little bit and change the aspect ratio of the fireplace. What a joke.
Note: The following is not about autos.
I just had an email from a friend that makes a good counter argument for AI for some applications. He is skilled in many areas of software, including AI. He is currently on vacation from Dresden in Hawaii on Oahu. He had Claude create a one day itinerary for a visit to Maui for a trip up Haleakala. He spent less than 30 minutes generating prompts to create the trip plan. He sent me the 15(!) page plan for my comments, since I’ve been to Maui many times because my wife grew up there. The plan was impressive and I only had a few brief comments.
I’ve discussed AI with him. He said that he could replace more than a few novice programmers with the capability of AI. His concern is how will his skill be passed on, if there are no new novice programmers. I use the term ‘programmers’ with some trepidation, since that minimizes what he does. I worked in materials (mostly thin-film deposition) during my career. For that field AI may be a help, but would require that the published literature, mostly behind paywalls, be scraped.
Now there are more papers being published in arXiv.org, a free distribution and open-access archive of papers that have not been peer reviewed.
Sorry for being long-winded about non-auto related postings.