As a former translator with two rock-solid university degrees in the subject, there was never a universe in which I would not talk about Twitter’s new autotranslation feature turning the tamest things into hardcore pornography.
Elon Musk’s AI chatbot Grok has long garnered a reputation for experiencing horrifically racist meltdowns, enabling child abuse, and doxxing users’ home addresses.
It should come as no surprise, then, that its supposed “translation” is a piece of work, too. In April, the almost-trillionaire’s social media platform X instated automatic AI translations for all of its users — and the results certainly speak for themselves.
As writer and author Parker Molloy pointed out in a recent post on Bluesky, the Grok feature is “taking some interesting liberties” with people’s otherwise sincere posts.
Screenshots show how Grok completely botched translations by coming up with shocking and decidedly NSFW AI hallucinations.
↫ Victor Tangermann at Futurism
The sloppy translations this garbage software comes up with are honestly quite hilarious when taken in isolation. It’s adding translations that are straight-up hardcore pornography descriptions to entirely tame material that has absolutely nothing to do with pornography. The description of a video of some guy making coffee is translated into “man masturbates and jerks off to his own coffee during commercial flight”.
We all know how this happened. There’s a lot of pornography on the internet, and Grok being the worst autocomplete among autocompletes, it was probably fed a lot of pornography, without any limitations or guardrails. The end result is obvious: some random videogame video is now a “cumshot video with my stepmom”. It would be absolutely hilarious if it wasn’t horribly dangerous.
I’ve explained countless times that “AI”-based translations are going to get people killed – probably already have, but we just don’t realise it yet – and it’s not hard to see how a slopmachine turning innocuous things into hardcore pornography can do just that. There are countless places in the world where a woman unknowingly sending a pornographic message to her parents or whatever can get her hurt – or worse. I hadn’t even considered this particular way “AI” translations could get people hurt.
Sadly, we will most likely never know the full extent to which “AI” translations will get people hurt and killed. When your grandmother takes her medicine in the wrong way because the “AI”-translated leaflet was unclear or downright wrong, and she ends up in the hospital because of it, will you ever find out what caused it?
First, I want to say that I sincerely sympathize with you. Having two degrees and not being able to live comfortably off of that is infuriating and unfair.
I also don’t really use X or Grok, so dunno.
But as someone who only speaks two languages, machine translation is a huge unlock in my personal life, even if deeply flawed. It beats having nothing. I know it’s a selfish thing to say, but there is a an upside to this for many people that cannot be ignored.
We may be better off without AI, but the genie is out of the bottle, and there is no putting it back in. No amount of quotes around it will box it in.
drstorm,
I’m still fascinated by the “growth spurt” of AI, and trying to frame why some professions are embracing it, while others look at it as the incarnation of evil.
As a developer for example, anything that makes me type less keys to do more feels like a positive. However many in creative endeavors, think the opposite. Forget embracing the generative tools, they treat it like dirt.
Why?
I think I can explain that: in every team I have worked with, the usual 20:80 rules apply. 20% of the people do 80% of the work and need to carry the dead weight along because of some “social contract”. I am very sure that deep in their heart, both sides know about that.
Now, the news tools accelerate the 20% and make the 80% just obsolete. If you own the company/IP you can develop and produce faster. Sketch, code, documentation, commercial offer. Done in a few days (what took months before) without depending on anyone. I want to work Sunday night: go ahead. I have something else to do Monday morning, no problem either. I have never been so productive in my life like over the last 12 months (Claude 4.6 arrived).
At the same time, I did also some paid programming for some other firms before(JSQLParser/JSQLTranspiler) — they don’t need me anymore, Claude 4.6 killed it and this is completely fine (if your life does not depend on it).
Andreas Reichel,
I don’t see changes happening too quickly, but year after year the momentum will keep building and be impossible to stop. Gray bears will still have their jobs, but companies might significantly cut back on new employees. Long term I think job displacement is going to create a lot of hardship for new grads especially. More than ever one will need connections for the privilege of a job.
Even customer facing jobs are going away. I really dislike it but stores all around me are replacing cashiers with self service kiosks, I say my first gas station where I had to use a kiosk. As this is happening we’ve been facing large price hikes… We’re loosing service and paying more. Soon enough I expect those jobs will be gone and more behind the scenes tasks will be automated as well. I don’t know how people can make a living as automation kills off more jobs. This is going to turn into a huge problem for society.
Absolutely in agreement, I share this concern. And I would love it to see this educated discussion here (which is technical and social) instead blind activism and agitation.
Yet this has nothing to do with AI/LLM, but all with the asset inflation caused by helicopter money since 2007 and your economy who spends more on military interventions than rest of the world together while fly over states barely have power lines.
Andreas Reichel,
Sure, I don’t claim that price inflation is caused by AI. However the moral justification for why society accepts such a tradeoff is that goods/services are more affordable. I have no doubt automating jobs is cheaper for employers, but as a consumer it’s incredibly frustrating to see that when these changes are brought they do not correspond to lower prices. Someone is benefiting from automation, but it’s not the working class who are getting boned on both sides of the equation: jobs going away and prices going up.
The world is going to start running out of oil in 20-30 years. Completely economically depleted in 40, so gas prices are going to go up regardless. This Hormuz crisis is just a small taste of the next few decades, and the world isn’t really in a hurry to get away from using it.
Oh that one’s easy. “Creative endeavors” generally don’t pay as much, so people who get into those tend more overwhelmingly to be people who actually like creating art and like the process of creating art. And so of course they’re going to treat a “toolset” that tries to do the creating for them as dirt, because such “tools” miss the entire point of why they are doing it. They also often care more about ethics than your average IT person.
So there’s not much of a mystery here.
Similarly, I got into IT and computers when I was very young because I enjoyed it. It caused a fair bit of friction when I got into software engineering at university and discovered how many didn’t, how many lacked most of the curiosity that guided me there, and only seemed to be there in expectation of a higher income bracket. So I’m not particularly surprised by how many in this field will jump at trying to have generative “AI” do their work for them.
For my part I still haven’t touched a coding agent, or even a chatbot. I consider the primary part of my work to be building a deep understanding of the systems I’m supposed to work with, administrate and modify. And I enjoy doing that. So even if I didn’t have a host of ethical issues with these kinds of “tools” (which I do), I honestly still don’t see any significant ways they could help me, at least not in the way I tend to approach my work.
Book Squirrel,
I felt the same way. I had a genuine interest in CS. But I’ve found that the vast majority of employers who would hire CS employees have work that has little to do with CS skills. There also seems to be a stereotype that everyone in tech is doing great financially, but outside of big tech hubs salaries drop off very quickly and even in tech people can struggle to find rewarding work and I know more people in this boat than not. I’ve learned that people have very different insights into the industry depending on the quality of their job network. I personally made the mistake of focusing too much on my CS passion when I should have been focusing nearly 100% on my professional network instead. Connections are more important than anything else.
I do experiment with them but I haven’t used LLMs to do my actual coding work. I know people who do though. Right now I have a project where we’re looking at using LLMs to extract structured information from completely non-structured call transcripts. So far I am impressed with the LLM’s ability to identify what people are talking about. I might have tried creating some complex regex based solution with extremely fuzzy matching logic, but given that the data is completely unstructured and every interaction is unique, the scope of the problem seems to align with LLM very well. The alternative is to have people scraping the data in a mechanical turk kind of way, but the client asked me to find a way to automate the task.
I am also absolutely flabbergasted how and why someone can spend so much time and resources on telling me how much he despises certain products or companies. I am not using X or Grok either. I also would never drive an American car or eat Marmite (or whatever this organic glue is called the brits put on their card board) — but I feel zero need to tell any of those users, how stupid they are.
I don’t think Thom sees AI as a product, but as a theft of intelectual property. When you frame it like that, it becomes about the top 0.01% taking all our stuff and selling it back to us.
While I think there is a real problem here we as a society need to deal with to reward creativity and protect human dignity, I personally do not see AI output as theft.
Instead, I see it as an entity that learns and is able to reproduce what it has seen, very much like the many people who also don’t have a creative bone in them. When I ask AI to solve a particular problem I have, it uses the knowledge that comes from the work of many people who dealt with similar problems, sure, but how is that different than me doing it? I too learned from others.
> Instead, I see it as an entity that learns and is able to reproduce what it has seen, very much like the many people who also don’t have a creative bone in them.
I do understand where you come from, but even on the “creative bone” I am not too certain anymore: Take any kind of ugly UI, screenshot it, feed it into Claude and ask it to mock up a new modern take on it. The results are amazing and scary at the same time. We can label this as a kind of mechanic creativity — but so was any principle of art (at least until Dali and Picasso came along).
All I was saying is that Claude is unlikely to come up with a type of design you never saw before that is also brilliant.
Btw, to quote here you don’t use > like in Markdown. You use a q tag, but instead of angle brackets, use [/].
drstorm,
I feel the same way. Originality can be a virtue, but I think genuine instances of this are rare IRL. Almost everything we do and think have already been done and thought before. And this isn’t even our fault, our brains just do it naturally as a biproduct of our social reality. Throughout our lives, from infants,, toddlers, schooled aged, and so on, we train our brains on pre-existing human works, which is how we become proficient ourselves. I find it very difficult resolve the contradiction of human training being normal, but AI training being “theft”.
Of course sentient AGI doesn’t yet exist, but this may happen in our lifetimes and then I do wonder if these double standards will be applied. It kind of reminds me of Data’s trial in Star Trek.
“Star Trek: The Next Generation – Sentient Being”
https://www.youtube.com/watch?v=vjuQRCG_sUw
I want to add that even the US Army denies applicants with less than 80 IQ — because they impose more danger to their own troop than to the enemy. Now, go and find out the portion of the common population with less than 80 IQ!
Yet, we would not deny them the “sentient” attribution, when at the same time even a weak model can do so much more already (just w/o a body around it). I wonder, how we will finally resolve this dilemma.
(Fun fact: I was evaluated and excluded from the draft as well 🙂 )
Absolutely machine translation is exceptionally useful. The problem is not machine translation itself, but the people who blindly trust it and make decisions or take actions based on its output.
At most it should be used to get a general idea of the meaning of non critical information, which is fine if you’re just casually browsing news or random personal blogs etc. For anything important or that you’re going to take further action based on you should be consulting a reliable source that you trust.
Grok continues to have problems that will impact its viability. Caught between the best and the cheapest I doubt it will grow legs.
Grok isn’t perfect but it’s the most useful AI for everyday queries and research. ChatGPT is on par I’d say. Gemini is completely useless and just makes up stuff all the time – despite having all the information from Google’s web crawler to hand. Perplexity is also really bad for research even though it’s supposedly it’s raison d’etre. In my experience, Grok and maybe ChatGPT are the only ones that know what’s going on and you can ask it a question about something that happened yesterday or last week, rather than having a model with knowledge cutoff in 2025 and then a bad web search bolted on to it. Claude supposedly has the best models but outside of coding I wasn’t happy with it. Mistral is just bad, even though I wish the one and only European AI company would be competitive.
So overall, as a non-coder, Grok and ChatGPT are the only ones I’d consider paying for.
Entry level translators, junior developers… AI has killed both. It’s something like 5x productivity gain, but only 2x review time by people with senior level skills. Eventually this will create an immense skilled labor shortage similar to the baby boomers retiring and then being hired back at multiple times their salary as there were no juniors to move up to those “reviewer” positions.