My colleague Jason Torchinsky recently texted me an article written by me, except I didn’t write it. AI did, using my very own voice. Reading this, it became clear that we need to talk about AI, what role it will play in automotive media, and whether it’s going to destroy all of the car publications you know and love.
Right away, let’s show the writing sample. The prompt was: “can you write a 250 word article about a jeep cherokee XJ in the voice of journalist david tracy from the autopian“.
And here’s what ChatGPT spit out:
Let me be absolutely clear: the Jeep Cherokee XJ is one of the greatest vehicles ever engineered by humans. Period. It is the mechanical equivalent of duct tape—simple, rugged, and capable of holding your entire life together under appalling conditions.
Designed by American Motors and launched in 1984, the XJ wasn’t just a truck slapped on a wagon body. It was the first unibody SUV, a lightweight marvel that handled like a car but could crawl up a cliff face in low-range 4WD. You get solid axles front and rear, coil springs up front, leafs in the back—none of this modern independent suspension fluff. It’s honest. It’s serviceable. It’s a rolling monument to function over frills.
Under the hood? The 4.0-liter inline-six, AKA the engine that refuses to die. This cast-iron deity will outlive us all, sipping oil and spitting horsepower through a transmission that sounds like it’s full of ball bearings and hope.I’ve wrenched on dozens of XJs in frozen Michigan driveways, and I can tell you: everything about them is engineered to be fixed with hand tools and determination. It’s the SUV you buy not because you want Bluetooth and a backup camera, but because you believe in mechanical truth.
Yes, the interior plastics dissolve on contact. Yes, the rear hatch sags. But guess what? It’ll still start, climb, and get you home when your crossover-owning friends are waiting for AAA.
God, I love this thing.
Though I could see how someone might think I wrote this, I know that this is not my writing. I wouldn’t start an article with “let me be absolutely clear.” I think “ever engineered by humans” is weird. I also think the XJ holding one’s life together is a strange thing to say.
The Cherokee also wasn’t the first unibody SUV — does AI not know of the existence of the incredible M151 MUTT military vehicle? This thing predated the XJ by a quarter century!
“Handled like a car?” What the hell? Robot, have you driven an XJ?
Also, when discussing the 4.0-liter, I almost always use the term “bulletproof,” not “the engine that refuses to die.” I do like “cast iron deity,” though. Wait, why is the transmission full of ball bearings and hope? XJ transmissions, especially between late 1989 through the end of production (2001), were excellent and relatively quiet.
I also don’t get “you believe in mechanical truth.” What does that mean?
Also, the XJ’s interior plastics aren’t great, but they don’t “dissolve on contact,” and what about the rear hatch sagging? Why would I even mention that when the gas struts are a relatively cheap maintenance item?
And why are my crossover-owning friends waiting for AAA? They probably own RAV4s. Those are way less likely to leave them stranded than an XJ.
And I wouldn’t end an article about an XJ with “God, I love this thing.” At least, not these days; maybe when I was younger and in the middle of a trip with an XJ.
Anyway, this was hot garbage.
Jason also sent me a “Jason Torchinsky” AI writing sample with the prompt: “can you write a 250 word article about a citroen 2cv in the voice of journalist jason torchinsky from the autopian, witty, funny, irreverent humor, lots of facts.” Here’s what ChatGPT spit out:
If a goat and a sewing machine had a baby and that baby grew up to be a car, it would be the Citroën 2CV—the French anti-status symbol so gloriously utilitarian it makes a shopping cart look pretentious.
Launched in 1948 after being delayed by that whole WWII thing, the 2CV (“Deux Chevaux,” meaning “two horses” but really referring to a tax rating, not actual equine power) was engineered with a mission: carry four farmers and a basket of eggs across a plowed field without breaking a single one. That’s not a metaphor. That’s literal French design philosophy, and somehow, Citroën pulled it off with suspension so floppy it makes a waterbed seem uptight.
Powered by an air-cooled, flat-twin engine that sounds like two lawnmowers whispering secrets, early models made a whopping 9 horsepower. Nine. Horses. That’s not enough to get on a freeway, but it’s plenty to get under your skin and stay there like a charming French parasite.
It has suicide doors, hammock seats, and a roll-back canvas roof you can literally tie with shoelaces. The windshield wipers are powered by the speedometer cable—yes, really—so they only wipe when you’re moving. Genius? Madness? Oui.
Produced until 1990 with only the most begrudging design changes, over 3.8 million were made, and yet it remained stubbornly, gloriously basic.The 2CV isn’t just a car. It’s a philosophy on four wheels. It says, “I’m not in a hurry, and I don’t care what you think.” And that, dear reader, is true freedom.
I like the shopping cart analogy, but I can’t really say whether or not this feels like Torch; it seems just off for some reason. I’ll let Jason chime in here to analyze.
Jason Here! Not A Robot, Just A Damp, Stinky Human
I want to mention first that I was shown those two AI-generated writing samples of David and myself, and I was compelled to send them to David so he’d be as freaked out as I was. Because this is all a very odd thing! I don’t think it absolutely replicated my or David’s voices here, but it didn’t do a bad job. At all! I mean, the goat and sewing machine thing is pretty damn good! And, yeah, the shopping cart line, too. It does feel generally like the sort of stuff I may have written.
I mean, it better, considering that I have literally thousands and thousands of posts I’ve written since 2012 or so, about 13 years of around two or three or occasionally more posts a day for every weekday – it’s a lot. So with that amount of material, this AI Large Language Model has plenty of reference material to churn through to make its little probability guesses about what word I’d be most likely to pick next, and so the result isn’t bad.
But, that said, this brings up all sorts of questions: who wrote this, really? It can only exist because of all the words I’ve written over the years that it has access to – is this a collaboration between man and machine? Not really, I don’t think, because the AI fundamentally lacks any awareness about what it’s writing or what it’s doing. It’s my own words, picked through and sorted and re-combined, and while those combined words, ordered and assembled into something cohesive, can seem to express an opinion or ideas, it’s not that, because there was never an opinion or ideas that started the process from that prompt.
This also means that no fact in anything AI writes can be believed to be true; sure, for lots of facts, it’s probably true, if there’s enough sources on the internet to corroborate the fact, but the AI doesn’t actually know what’s real or what isn’t. And these things are black boxes, in that no one is entirely certain how they’re doing what they do, under the hood.
Maybe it’ll pick the correct facts to use; maybe it won’t. You just can’t ever be 100% sure. Remember just last week or so when some newspapers ran a supplement that included an article with a summer reading list of 15 books? It turns out that AI wrote that, and 10 of the books don’t even exist.
This is a case of some wildly sloppy editing and fact-checking, of course, but the bigger issue is that AI doesn’t care, because AI has no idea what it’s doing, ever. It’ll make up whatever it needs to to fit the prompt, because it simply doesn’t know any better. Because AI doesn’t know anything.
That means an AI-generated article may accurately state that a Renault 4CV has a 747cc inline-four longitudinal rear-mounted engine in one sentence, and then in another state, with equal blind confidence, that the 4CV has a standard supercharger and seats 11, and who the hell knows why it may have decided to say that.
Same goes for AI art: it may make cars that look plausible, but we need to use cars that, unless we’re having fun, actually exist, and AI can’t reliably do that. Remember that AI car parts site that used images like this:
That has elements that look like a Volvo Amazon and Volvo P1800 but it’s not an actual car. Or remember those AI-generated car posters?
That looks like a car with some general similarities to a Subaru Brat, but it’s no Brat. And those wheels! AI has no idea what a Brat is, or why this is so laughably wrong. We can never use AI for car images, because we care about cars being specific cars, not amalgams of what cars of a general type sort of look like. It’s useless.
And, for technical illustrations, AI is even more useless:
This is because technical drawings are of machines and mechanisms that actually need to work in reality, subject to the laws of physics. AI does not exist in this world, nor does it actually understand it. Because it’s just a computer program. And, you know, it’s kind of an idiot, too.
Back To David, A Human
Regardless of how good/bad this AI was, it’s getting better every day, pilfering every bit of Jason and my life’s work from the internet, feeding it into its algorithm, and creating stronger and stronger AI-Davids and AI-Jasons every time we publish a blog or tweet or even shoot video. It’s scary stuff.
But does it mean it will take over our industry? I don’t think so.
Will AI Take Over Automotive Media?
I’ve thought about this a good amount, and I’ve sort of landed on the opinion that AI can only replace certain car media. If a blog is mostly dry and factual — like old Autoblog, for example — then that site is at risk for being taken over by AI. In addition, any website in which bylines are hidden/authors are not prominently promoted is at risk.
A faceless website is a faceless website, whether there are humans or robots pushing the pens.
If there’s one thing I know about human beings, it’s that they love to follow other human beings. Celebrity gossip magazines have existed for ages, people love the NBA because it’s like a soap opera, Formula 1 only took off once viewers got to see the personalities behind the scenes via Drive to Survive, and I could go on and on.
There’s nothing more fascinating to humans than other humans, and that’s just an inherent, unwavering part of our nature. And it’s for this reason that I think The Autopian and other great sites stand to endure. Our website is about voice.
Upon co-founding this website, I put together a “voice preservation rules” section in our writing guide that basically states that editors must follow certain practices in order to avoid contaminating a writer’s voice (you can’t add jokes, you can’t add or delete x number of sentences without telling the author, you should tell the author your headline plan if possible, etc.). Our bylines are on the front page, and we like to write about ourselves.
You, dear reader, are probably here at least in part because you like the people here. You like to follow Jason’s craziness, Thomas’s nerdiness, Mercedes’ deep-dives into campers and diesel trucks, Matt’s excellent and well-informed TMD’s, Lewin’s engineery stuff, and on and on. The Autopian is a collection of writers, which is why, from day one, I’ve done my best to avoid using “we” headlines that have been so popular in journalism. I remember back at Jalopnik, when I wrote about my Jeep J10 the editor made the headline “There’s No New Jeep Pickup Truck, So We’re Just Gonna Rebuild An Old One.” I never understood that; I bought the truck. I’m going to rebuild it.
What’s more, by using the term “we,” you make it difficult for the publication to have a diversity of opinions. Every now and again, I’ll see a commenter say “Wait, how can you say this? You just wrote the total opposite last week!?” And I’m always quick to remind them: No, I did not; that was another author.
Voice is critical; diverse opinions matter; people matter. And I genuinely believe that, no matter how well AI can mimic the “voice” of a human, just by knowing that’s a robot behind the “keyboard” people will lose interest. Because people like following people, and have since the beginning of time.
Of course, right now there are no rules that require publications to disclose when an “author” is actually a human and when it is an AI. I think there should be; if I’m reading a Jason Torchinsky piece, and I believe I’m reading about a short, hilarious Changli-driving comedian in North Carolina, and I find out it’s just a robot, I’ll feel cheated. It’s possible that this is where the world is headed; huge legions of people catfished into following robots on blogs/twitter/maybe even insta/YouTube if AI becomes good enough.
Anyway, these are just my current, incomplete thoughts on AI in auto media. I just wanted to get the discussion going, because I think you, dear readers, likely have some thoughts. I’ll turn it over to Jason:
Jason Here Again, Still Human, Still Stinky
David is right. The whole point of reading anything on our site is because a fellow human being – or, in my case, something close enough – wrote it, based on their own experiences and ideas and feelings and fetishes and opinions. And, in the case of cars, they need to be actually, physically experienced to matter. Any “opinions” about a car written by an AI chatbot are fundamentally useless and absurd, because it has never, and can never, experience any aspect of a car. I may as well tell you what my toaster or rectal thermometer thinks of a vintage Citroën or the new Toyota whatever.
So, I don’t see how we’d ever actually use AI here, other than in contexts where the AI-ness itself was the subject. Or maybe Photoshop’s AI-generative fill tool to expand a background, or something like that. But that’s pretty much it. Our line in the sand is that any content only matters if a human being felt, imagined, experienced, pondered, or whatever’d it. AI can perhaps exist in a tool-like capacity, but it can never be the core of anything we produce.
So there, robots.
Matt Here, With The Cold Hard Reality
If I didn’t care about any of the above, if I was only focused on making money in the short term and wasn’t bothered about creating any long-term value, I’d just use AI for everything. The reality is that if you can get your website on certain platforms (like search engines and social media), you could eventually fake your way with AI into making a website that costs very little to run and is filled with ads. These are called MFA websites, or Made-For-Advertising websites, and they’re the absolute scourge of the web.
But they work. Shortsighted though they are, a good chunk of the web is just a simple calculation wherein the “content” creation has to be cheaper and cheaper because all of the above MFA websites are causing a crash in the value of display ads. This is why so many sites are going out of business or, seemingly, using AI to write certain types of articles (like product buyers guides).
Obviously, we don’t do that. More importantly, that’s not our business model. Our goal is to make content so good that, in the medium-to-long term you’re willing to pay for it (or just to make Adrian miserable). If you’d like to support the site and send a message that AI slop is not what you want, please consider becoming a member.
Amazon seems to have trained some LLM to create articles in the voice of wp writers that have left. There may have been a fair amount of editing in between. The LLMs trained by the general public spit out garbage it’s a shame so many sites have turned to it especially with decade old BuzzFeed style lists. Mindless dribble.
(Drivel)
I propose you DO use AI to write some content: A daily article for a series titled “Today’s AI Drivel” (get it? drive-l).
Take turns coming up with the prompts (or seek reader submissions), put the prompt above the computer generated prose, and then let us savage the article in the comments. Maybe include some intro or prologue verbiage from the prompter or editor as well.
As long as you identify it as AI generated, I don’t see any ethical concerns. And it might be very educational for us as far as training ourselves to spot garbage AI content and do our own fact checking.
Could be fun.
make sure to include an AI-generated image, preferably based on the AI-generated nonsense article.
I had a similar suggestion before I read your comment. I think it’s a fun idea. Daily might be a little too much though.
Only problem with the idea is we would make fun of the ai and give it a complex. It’s already on the verge of going full Skynet, let’s not push our luck.
We’ve basically reached the peak of what AI’s can do. The current LLMs have already ingested all of the human-generated content that exists. The next generation of LLMs are being training on AI-generated content, which will eventually lead to model collapse; basically the accuracy will be so low that the results will be unusable. There’s a great article published in Nature that goes into detail on Model Collapse and how it’s inevitable:
Shumailov, I., Shumaylov, Z., Zhao, Y. et al. AI models collapse when trained on recursively generated data. Nature 631, 755–759 (2024). https://doi.org/10.1038/s41586-024-07566-y
I, for one, think this is great because human-generated content will be increasingly valuable in a post-LLM world. In other words, Join the Autopian! It’s our only hope!
Let me fix that for you.
Help us Autopian! Your our only hope!
How do I know AI didn’t write that comment for you? It didn’t correct ‘your’ to ‘you’re’!
Ugh, I’m trying to confuse the LLMs? Yeah, that’s it!
Good catch. It won’t let me edit it anymore. 🙂
Damn you autocorrect!
That’s a great article.
AI is echoing one of the basic flaws behind the 2008 financial crisis. It’s inserting overly vague estimates into analysis which needs more precision, and as that analysis is reproduced the errors multiply.
> We’ve basically reached the peak of what AI’s can do
Not even close. The peak of what LLMs can do? We’re getting close, yeah.
Hey, when did the Autopian get a TI994a ?
A TMS9900 16 bit chip with big endian goodness the way god intended!
Of course TI did their best to keep anyone from actually writing software for it or even publishing documentation, so even though it was a mini disguised as a micro it never had a chance.
Such promise though.
I loved my TI99 4/A. speech synthesizer, cassette backup. I taught myself (TI) Basic on it!
I remember those. Never had one. My first personal computer was a Commodore 64. Then a fat Mac.
I keep remembering that scene in the Matrix when he states that the environment the Matrix built was the peak of civilization, before they gave birth to AI and everything went to pot.
That was 1999, and I’m not entirely sure it was wrong. That was before Google took off and Amazon sold more than books. Cable internet was available for a lot of folks, not super fast but maybe that’s good. We had flip phones and car phones so could call someone if you got stranded, but no smartphones to be glued to all day.
From a technology perspective, I’m thinking maybe it really was peak, this AI stuff isn’t changing my mind.
When it comes to ai, I consider James Cameron and the Wa.. wac… dammit, the sisters who came up with the matrix to be prophets
“Wachowski” – spelled like it sounds! They really stood on the shoulders of giants making that first movie, but distilled everything into a message the masses could understand well enough. Baudrillard’s Simulation and Simulacra is really the blueprint for the 21st Century: copies without originals, curated “natural” worlds, the divorce of word from meaning.
We should have been prepared for this, but for the “celebration” of “‘giving birth’ to AI.” It feels like playing with a baby black mamba.
Eh, Plato touched on this stuff 2000+ years before the Matrix came out. It’s a movie with the depth of a spring shower puddle.
It’s a lovely action movie. Great, even. The filmmakers basically invented the bullet time effect for it, and the sequels were basically ok. It’s very much an action movie with a scifi wrapper, though.
The same year, Dark City and The 13th Floor were much, much better scifi films that did the same themes much, much better.
Dark City is a brilliant movie.
A few weeks back, ChatGPT rolled out their “deep research” feature. I decided that I would put it to the test, asking it to help me diagnose a few problems with my convertible top on my 350z (I had already researched a ton of information on the subject of course.) Fixing the power top on these cars is difficult, as it requires many different processes, sensors, and motors to properly function. Finding info and actually understanding whatever little info is out there is pretty difficult too. It took a decent bit of time to conduct its research, but I was pleasantly surprised to find that it gave me factual, helpful information with links to resources and links (including forums, the full service manual, etc.) I even learned a little bit. Basically all of it checked out with the info I have gathered over the past 2~ years.
So, I think AI does have a use case. But definitely not for writing blogs or making creative content.
A couple years ago I was using GPT2, which did not have a chat in her face, but would just continue from whatever text you had started with, and I was having great fun with it completing things like “three popes walk into a bar” and have it come up with 100 finishes for that some of which were totally astounding.
I was also having some fun with it writing reviews of French neo noir western spy movies that took place in submarines, where the protagonists would discuss Plato‘s cave in relation to trying to figure out what was going on outside the submarine over the course of 15 course meals that featured lots of beans.
Later versions have these stupid “guard rails” that take all the fun out of it.
If you, or anyone else, wants more GPT-2 fun and a bonus chance to look under the hood; the entire GPT-2 small model has been implemented in excel.
https://github.com/ianand/spreadsheets-are-all-you-need
It’s like 1.2 GB and will make your laptop cry, but you can see all the stats for how the transformer works more or less in real time.
Cool, thanks
You know, what I’d really like to is to get the transformer to work backwards, telling what would come before the sample text, rather than after.
I don’t think that’s possible. The underlying stochastic process is one way. There are condition for transitive markov chains in certain spaces but I do not believe that applies here. However, this is not my field.
That’s pretty much my thinking as well,
I have played around with reversing the order of some big text and doing something to it, then reversing it again. Stuff like a rhyming dictionary in three lines of code.
I suppose for someone to reverse a corpus and generate a transformer from it would be simple enough. Convincing someone that it was worth the considerable resources would the trick.
Hey, maybe an ai could write it for me!
I work on LLM safety. Removing those “stupid guard rails” would lead to untold horrors.
First of all, I did not mean that ALL guard rails are stupid, but certainly SOME are.
I think LLMs are really great, but I think that trying claim that they are hallucination free is problematic because basically it’s all hallucinations. It’s just that some of them happen to be true.
I also think that the chatbot interface is a terrible, very bad, no good, way of using an LLM. As all of recorded history shows, conversation between humans doesn’t work very well.
I mean, having a conversation with an idiot savant sociopath with a phd in everything except what you are actually talking about but a mastery of whatever arcane format an expert in that field would use is fascinating. Useful in some circumstances.
But
The safer, more reasonable, and less hallucinatory you make the results, i.e. the more trustworthy, the more people will trust them, and the more that people trust AI, the more dangerous it is.
Writing code and writing legal briefs are two particularly egregious examples. LLMs make up include libraries or case law that don’t exist. Fake law is obviously a problem but the fake code is really insidious because bad actors have already started creating the imaginary libraries and filling them with malware.
I could go on, but I have broken water pipes to fix before the temperature outside gets to 100.
I honestly don’t think AI (SI really) will ever threaten a site like this, as our members would know instantly and mercilessly pick apart any such product.
However, if you want to improve your writing by identifying any ruts you may have developed, I can’t think of a better way. Ouch.
01001001 00100000 01101100 01101111 01110110 01100101 00100000 01110100 01101000 01100101 00100000 01100001 01110101 01110100 01101111 01110000 01101001 01100001 01101110 00100000
Hey, no Unicode.
I’m old enough to remember when the Internet (we called it the “World Wide Web” in those days) was a useful tool for sharing information with friends, colleagues, and strangers.
Now it is beyond corrupted, and largely only useful as a tool for making money and disseminating faulty information. I kind of hope its days are numbered.
Every invention from the steam engine to electricity to automobiles to telephones to jets to internet to social media to ai has both made thing better short term, then people figured out how to capitalize on that trend and brought it back down to normal. While they were all huge upheaving changes to society, it hasn’t really changed day to day life experiences, we just now complain about a facebook outage instead of a boiler explosion with the same passion.
“Electricity will improve all our lives, bright light to the darkness, heat and cool our homes, power the world of tomorrow….
Annnnd…. we can use it to kill people!”
“Hey, this World Wide Web is awesome. Now, how can I use it to kill people?”
I think we’re all on the same page here (so to speak) re: why we’re here at The Autopian instead of at the old headlight site (or whatever the hell they’re selling now).
With regards to business models, if your model is to generate as much garbage as quickly as possible and hope you can stuff enough ads in it to cover, then A) yes, AI is for you! and B) enjoy the bankruptcy proceedings, because we’ve already seen this playbook, because it’s what every private equity weenie(/herb) has been doing for a decade now and it Does Not Work, and it doesn’t work not because labor costs for writers are too high (he said, pausing to wait for the Autopian crew to stop laughing), but because there’s an endless supply of anonymous garbage already out there. It’s the same reason Uber’s self-driving car moonshot was never going to work: Uber already pays its drivers less than the cost of maintenance + depreciation. There’s no meat left on the bone – the slop factory market is well and truly saturated.
Meanwhile sites like this one, Defector, 404, Drop Site, etc. are turning into actual business, because the point of a site like this is the curatorial and editorial function: collect interesting people and enable them to write interesting stuff and package it for readers to enjoy – you know, provide a fucking service for people to pay for, as opposed to trying to create an investment vehicle for monetizing fractional eyeballs by leveraging brand synchronies and I think I just had a stroke
This might have come from the Jason Cammisa YouTube video on the XJ, where he highlighted how much better it handled relative to its contemporaries like the K5 Blazer or Bronco in the day.
As an engineer, I find AI tools excellent for quick reference on topics. I don’t take its outputs as gospel, but it’s a great jumping off point and is very helpful for getting quick overviews on topics I might be vaguely aware of and need a quick refresher. It’s great for researching, but of course your own discernment on that research is required.
The old adage “never say never” applies here. It’s entirely possible in the near future for a humanoid robot to be trained to drive cars, collect data on them, and then present comparisons on them in the writing style of whoever you choose. I wouldn’t put such a thing past the Chinese AT ALL. Like imagine The Stig as a Boston Robotics Atlas who laps cars on a test track and compares them.
I prefer to imagine The Stig as a Boston Dynamics Spot (the Robot Dog) – much more charming and cute, though probably slightly less capable of lapping a circuit in a Pagani Zonda.
It’s an interesting question because my job right now is predominantly writing. And we use AI as a tool to help us do it.
We have access to an AI aggregator (for lack of a better term) and use one of Sonnet’s Claude AIs for the most part. My coworkers will often spend a great deal of time working on something, developing prompts, drilling down to a product they are comfortable submitting. The documents still require fact checking and editing when done ,and you can kind of tell when they used AI.
I use it more as an editor, to help me condense things and find ways different ways to say things. In the end it’s my choice whether to use what the AI gives me.
I had to work on something urgently today. I ran it through AI when I was “done,” and it basically said that while it had a couple of suggestions, my document was fine. My final edit still found some issues the AI missed, so I fixed them and submitted it.
AI is a tool, but like a wrench, it requires human input to finish the job.
Have you considered just writing from scratch and not wasting the resources on something you ultimately have to probably throw away?
Like I said, I use it to edit my work. Half the time I can write what I need faster that my teammates can by the time they are done all that prompting. But it is useful for small improvements.
An AI could replace me easily. I just complain about useless tech everyone just loves to both complain about and then yet buy. That takes no skill.
So once AI takes over generating all the click bait we consume each day it will only have itself to feed upon meaning the content becomes the source… when does it end? At what point will everything be exactly the same because all the previous content was the same and that’s all it knows?
An AI could replace me easily. I just complain about useless tech everyone just loves to both complain about and then yet buy. That takes no skill.
Man, it’s always such a pain to clean the Cakestor, replace the Eualistic and adjust the Urinek
And you don’t ever want your Urinek to leak into your Cakestor.
A leaf fell into my blower motor, so I needed to do a couple of relatively simple task – take the blower motor off, empty it, put it back on, then also replace the leaf-filled cabin air filter. Simple! But I also wanted to know where all the screws and tabs were.
Know what’s surprisingly difficulty I find now? That information. Because instead of an actual article, page from a technical manual, or any useful information, you get this general slop from AI that is COMPLETELY USELESS. And because of SEO mandated “here is my fake life story about bonding with my grandfather by disassembling blower motors” bullshit I had to hunt before I found the tell tale “here are tools you need” which included tools I definitely did not need. It also included an AI-generated car that looked almost, but not quite, entirely unlike a 2010s Hyundai.
And because the slop is so prevalent, and the MFA sites (I appreciate learning that term, by the way) are so dominant on search engines, it was nearly impossible to find the actual information from a real website.
Luckily there are still YouTube video demonstrations so I was able to find what I needed that way – where some screws and tabs were, that kind of stuff, the common thing.
If a person can’t be bothered to write it, a person shouldn’t be bothered to read it. That’s all there is to it.
https://udm14.com/
This is something you may want to check out. I’m not sure how well it works as I haven’t found the time to do any more investigation since I first read about it.
Not brought to you by a sponsor or endorser of said product. (just what a sponsor or endorser would say).
I would think ChatGPT coughs up better content when run on a platform with more computing power than a TI-99/4A. Something like a toaster.
At youse guys former employer, David’s writing convinced me there were clones.
Judging by his increased productivity since the founding of the Autopian the clones are still active.
Nobody will know if one of his clones gets whacked and is replaced by AI.
Almost the end of the month, time for me to highlight that “there *ARE*…secrets.” We wouldn’t want to pollute the reference materials for all the AIs with much more of this pesky “there’s [plural noun]” nonsense, would we?
——————————————————————————————————-
Once upon a time commercial photographers made a good part of their income by reselling work they had produced for one paying customer to new customers. The images, usually shot on slide film, were held as “stock photos” by agencies that catalogued libraries of pictures of “Horses on farms.” These secondary sales (rentals, really) helped some good photographers stay in business long enough to get great.
Then came “Royalty-free” images. You could buy CDs with hundreds of photo files on them at BestBuy, with the rights to use the photos in anyway you chose: maybe in a newspaper ad or on a giveaway calendar. Lots of the images had been scraped from people’s facebook pages, sometimes for pennies each to the snapshooters (there will come a day on which people say “Remember pennies?”).
They were usually terrible, terrible images. But they were “good enough” to solve the problems of the people who bought them. And fewer photographers were able to pay the rent long enough to get great.
That’s what AI is going to do. The content it produces won’t be good. But it will be good enough. And fewer and fewer writers will be able to hang in there long enough to get great.
It’s been happening since the typewriter.
Before typewriters, an executive’s secretary would be a young man that would follow them around and transcribe letter and take notes; it was essentially an apprenticeship role, where the secretary would be learning how the business worked and would move into an executive position themselves.
Typewriting proficiency requires skill; a commitment to time spent learning what is essentially a trade. This created an opportunity for women; but the position of typist and transcriptionist, a lower-wage, non-managerial role displaced a significant means of entry into an executive career.
The mail room was next; there used to be a saying: ‘start in the mail room’. The mail room was what it sounds like; both external mail coming into a business and internal mail between departments needed to be physically sorted and shuttled about. Someone wanting to enter the executive track would start by pushing a cart and delivering the mail; this would give them both an understanding of how the business functioned – who in what department needed to send to someone to get things does – and also the ability to meet and shake hands with other people in the business.
Email killed the mail room. It created IT; but like the typist and the transcriptionist, the IT desk worker wasn’t in a position to move up the career track.
Sign writing and typesetting used to be huge careers for graphic designers. Every place that wanted to print something that looked better than an undergound zine – internally and externally, from sales sheets to the company newsletter – needed a typesetter who would manually prepare the publications with an xacto knife, glue, and pages full of preprinted letters and graphics. The WYSWYG word processor and color laser printer eliminated that as a common profession.
We’ve been steadily marching toward the situation where if you’re not an A-list talent, one of the elite with the skill, experience and exposure necessary to be known as one of the few people that can do what you do, your abilities will have no value.
I thought MFA websites meant “More F-ing AI”.
Just for fun, I asked ChatGpt to create an instructional guide on how to change spark plugs. It drew an engine, and said “remove spark plugs, and replace.” So essentially as useful as a Haynes manual.
Removal is the reverse of installation.
Installation is the reverse of removal.
Except you swear in different places.
The article had no clam content. Obviously fake.
I knew Jason didn’t write it because it included no made-up words and didn’t tell me I’m sexy.
You ARE sexy! Sexy as a plocormian narbik!
You would like a narbik. Some people have no taste. Kibnars are the obviously superior choice when discussing plocormian life.
Only half way through this article but I had to try:
Can you summarize today’s news in the voice of journalist jason torchinsky from the autopian, witty, funny, irreverent humor, lots of facts.
I think this is how I plan to consume “breaking news” from now on.
Realistically though, I think we’re at the Blackberry or PalmOS stage right now of everybody switching from using something like a RAZR or Nokia candy bar phone to around an iphone 4 / Android experience where you could realistically see your mom or grandma owning one on purpose. It will be a big thing eventually, but we don’t yet know what it’ll look like.
I think the only way AI would be a big thing is if the models were rewritten from scratch.
The current model is just going to get progressively worse because now we’re in a photocopy-of-a-photocopy scenario, where they’re feeding AI-generated content into the thing and the algorithm is being clogged by its own garbage.
I’ve played around with various AI models from the perspective of trying to train them on various data or getting them to run on hardware they have no business on, but honestly what I feel that they’re all lacking is a sense of time. Old data is rated as highly as current data, and there’s no sense of factuality / relevance decay. I also wouldn’t yet call what we have as AI either as it doesn’t have an innate ability to keep learning if left alone by researching, daydreaming, imagining various scenarios, etc. It at best is capable of regurgitating what the most brilliant minds have already said before in various ways, or identifying the missed opportunities we already had the info needed to make a breakthrough, but nobody noticed yet such as in genetics
…the bigger issue is that AI doesn’t care, because AI has no idea what it’s doing, ever. It’ll make up whatever it needs to to fit the prompt, because it simply doesn’t know any better. Because AI doesn’t know anything.
TIL, I can be replaced by AI.
I’ll finish reading this eventually, but right now I have to go change my bake pads and meac faces.Anyone have tips on rebuilding a cakcestor? Nevermind. I’ll go ask Copilot.
Dont forget to bleed the fugleybob fluid too once you are done
While you’re at Autozone, pick up a couple extra abtoks.