> This is why I use AI for all my medical questions and doctors use AI to write software, and we both smirk at the quality the other person is getting from it.

There is an interesting third group emerging: People who acknowledge the quality problem, but think they can deal with it by applying more AI to the output.

This takes the form of people who spin up a lot of "agents" and give them personalities like security director or quality director (which are unnecessarily complex and maddeningly unpredictable ways to trigger an LLM session for doing a security review or a quality check pass).

It also includes the person who knows that their app is full of bugs, but thinks it's not a problem because they can have the AI fix the bugs as they show up. People in this class haven't encountered security breaches or data loss bugs yet. They think it's all about having Claude fix that div that isn't centered or handle that error code that shows up some times.

Well said. Everyone agrees AI can't do their job, so it ends up doing everyone else's.

I'm not sure how to formulate it yet but it seems there is some Peter Principle/Gell-Mann Effect corollary that is AI-related we can say here.

Perhaps: "AI rises to the level of its users' incompetence."

Or: "Confidence in AI output is inversely proportional to one's ability to verify it"

It seems to be a general principle: If AI is better than you at something, you use it. If AI is worse than you, you don't.

Each time the frontier models get better, I see another wave of AI doubters suddenly become believers. People say things like, "AI couldn't code last year, but now I use it for everything!" Interesting. Now we know how that the person who said this has the coding skills of a Claude Opus 4.5 or whenever the frontier was when they flipped.

Meanwhile, the rest of us keep using AI as simple tools, like the person in the article. I wonder how long it will take before computers can program better than me, and I flip too.

I was saying something like this a few years ago when people were getting first excited about ChatGPT. The gap has narrowed, but not by as much as people think.

AI produces output that is very convincing to a non-expert, and (dangerously), it's so good at looking like an expert, they might believe that it is an expert. But the moment you ask someone to use it for something they're an expert in themselves, the holes appear wide, consistent & obvious.

My favourite moment of seeing this in action was watching AI-worrier TV host/comedian Bill Maher. He has spent years talking about the dangers of AI taking everyone's jobs, destroying civilisation, ruining the economy, starting wars, "it's just getting better and better all the time", and so on. But one night he let slip a tell. "It's no good at writing jokes. Not yet, anyway". There you go, Bill... connect those dots...

There is real utility in it being a tool to help experts apply their expertise, as in this story where it speeds up some tasks to help the translator do part of the work, enhance their expertise, allow them to be more productive.

It's a better screwdriver, a better hammer, in the hands of somebody who knows what needs a screwdriver or a hammer. It doesn't replace them. It can't replace them. It's a tool that enhances the human, not an alternative.

I don't understand why this is not widely understood yet, but I'm sure it will in due course.

And I don't expect this to change. Even if the latest model scores 100% on every benchmark, all that really tells us is that it's now more productive/efficient than it was before at helping experts do that work, not that it can replace everyone in that category of work.

Reminded me of this post by EY. (You're making a different point about existing expertise, not LLM expertise, but I think it holds in general.)

Every month a new guy discovers LLMs; discovers a skill the current LLMs require to get good results; and writes about the future jobs that will always be available for smart people like HIM, that are SKILLED in using LLMs.

The next generation of AIs doesn't need his fancy prompt. The image model goes from needing to type in just the right set of weird words and cryptic sorcerous invocations, to most people being able to type in English what they want and get a pretty good result.

There are still tasks that require careful invocation. But they are a much smaller fraction of all the tasks people are trying to do, or you can get a bleh result without the elaborate invocation to get it really good. And to improve on the bleh result you need to be substantially more of an expert than back when the Guy was memorizing a rule about adding "trending on Artstation" to the image prompts, as would always require a human paid to do that.

Another generation of AIs comes out. The next generation of Clever Skills is obsolete. Image models just obey the instructions for compositing panels without mixing them up, and you don't need to be an expert to get them to do it right. Another human value-add is gone. A wider set of tasks require no human expert.

Now a new Guy notices LLMs have become useful in his field for the first time. He discovers they require SKILL to use CORRECTLY. He posts about how there will always be jobs for humans who are SKILLED in using LLMs like HIM.

But it is not an infinite cycle. It is not the same each time it repeats. Now the Guy is a highly paid programmer or a career mathematician in 2026, instead of a graphic artist in 2023.

In six months the models will no longer require his vaunted Skills.

And by then there will be another Guy.

But the process doesn't continue forever. The Guys are coming from fields that were harder and harder for AIs. The brief centaur eras are shorter and shorter.

Today it is writers who are laughing at how bad the LLMs are at their job, and who will perhaps soon be posting about how it takes Skill to get an LLM to do their job Correctly. But the models are coming faster, and the eras of kinds of human value-add in each field are shortening.

There is a point when you run out of Guys, either because the centaur eras are too short for people to develop SKILLs and post to Twitter about them; or because there are not lands left for AIs to conquer; or because ordinary people are not reassured by some Nobel laureate proclaiming there will always be jobs for Nobel laureates with the SKILLS to prompt robotized biology labs Correctly.

But we'll never run out of amateur economists who assert entirely without a brief contemporary example that there will always be jobs for humans skilled at operating AIs!

We'll run out of professional economists saying it when nobody is paid for that work anymore.

I guess we'll also run out of amateur economists when they're dead.

Source: https://x.com/allTheYud/status/2057136382817231151

As somebody who regularly reads translated works, including the occasional machine translation (MTL), they (MTL) suck. You got a hugely biased result, which you recognize.

Translation is hard. If you're familiar with reading translations from specific languages MTL works have a very specific smell to them, it's a bit hard to describe but it's there. A good translation is miles (kilometers, for those outside of the US) above MTL.

That's not to say that perhaps the latest LLMs will have better translation abilities, but that they are generally crap currently. Maybe they are fine for something very short, but absolutely not for longer content.

> Out of curiosity, I sic'd Claude Fable on the original French version of The Three Musketeers and told it to translate accurately, but also try and keep the same jovial tone as the original and do not censor anything. After it was done, I didn't read the entire output, but I did compare a few individual chapters between the Ellsworth translation and the Fable translation.

This isn’t a great test, because Claude almost certainly has multiple translations of The Three Musketeers in its training data.

> I did compare a few individual chapters between the Ellsworth translation and the Fable translation.

I'm pretty sure the Ellsworth translation is in the corpus. You basically instructed claude to regurgitate it.

The llms all have the more famous books memorized. You can trick them to recite them more or less word for word.

> As far as I could tell, nothing was substantially different from the Ellsworth translation and the Fable translation.

Crucially the full translation was part of ChatGPT’s training set. Recall is a pretty solved problem in machine learning.

How well does it translate a French novel published yesterday? Where neither the original novel nor any translations are in the training set yet? Or might not even exist!

I tried asking ChatGPT to translate a letter I wrote in Slovenian this weekend. It got the general gist but missed a lot of the nuance. Completely missed several of the little touches of tone where the right choice of synonym conveys a whole bunch of information.

> … considered one of the more accurate translations of the work.

I think you’re missing a big point of translating literary works. A purely “accurate”, phrase-by-phrase translation is often not very good; the actual literary style, the feeling and the allusions and references, often get lost that way. A good translation of literary work requires a lot of deliberate choices by the translator to deviate from literal translations in ways that convey the style of the original, or an extra layer of meaning that would be lost by an “accurate” translation of a phrase. Also, being consistent with these choices matters a lot, which OP claims LLMs are less good at.

You're very likely to get a somewhat circular reference; the key (for me) is that for 90% of the usages, "standard translation LLMs" are just fine - I still recommend a translator but they're more of a proof-reader for both languages, catching where something slipped through.

I see the difficulties more in other areas, such as technical translations, specialist books, user manuals, and translating UIs, where contextual information and a back and forth with the client is needed to clarify details, and (for user manuals and UIs) the translator has to put themselves in the mind of the user and has to consider the possible contexts and use cases.

> Again, I don't speak French so I cannot say for sure

This reminds me of the adage, that ChatGPT is really great at everything except my own work.

This is sort of missing the point-- people who dont deal with linguistics dont understand that there are multiple types of translation. There's word for word (which is what you're talking about) and sense for sense. If you let an LLM do all of your translation you're letting it interpret huge amounts of intent and context it doesnt (and probably cant) access. The ways in which this impacts the translation will forever be unknown to you and in the worst case lost forever.

So i guess in the end it just matters how important the work is.

LLMs are now being aggressively manipulated for propaganda purposes. Powerful people have realized that people believe LLMs, and treat them as authoritative sources of fact.

The number of lies, lies by omission, deceptive distortions, and fallacious argument tactics they generate is absurd, and increasing rapidly. Translation, when done as a service you are paid for, can't be relied on by propaganda bots.

An interesting counter-example: https://xcancel.com/ValerioCapraro/status/206506665753442336...

.

This moment is coming for software developers too

I wish more people had casual exposure to professional translators. Its a deeply important, vanishingly small segment of the human population and has been this way for at least many thousands of years. Also, it will continue to be!

My first rule—before doing anything else—when writing a sentence, is to check whether I could have removed the em dashes by re-ordering the elements.

Update: in case it’s not obvious, I am sorry. I could not help it.

Em dashes are really good actually and a standard stylistic choice for non-technical writing, particularly outside the US.

My writing used to be littered with them, but I now eschew the em in favour of en, as it has become too strong an anti-shibboleth.

I have also taken to being sloppier in my prose, as I’ve had stories rejected for being “written by AI” - when they’re shorts I wrote more than a decade ago. Reworked them to sound like a moron, accepted. Sigh.

> often used as an example of "acceptable" AI by people who are skeptics of LLMs and AI-generated art.

As one of such people, I think there is a nuance to it. AI is great when you’re translating something to yourself. But when translating things for others, more caution and human judgement is needed. Espesially when translating instruction manuals, where bad wording could cause someone to injure themself.

There are translators and there are translators. Translating legal/business documents is a completely different thing from translating movies/books/games.

I can confidently say that LLMs do a better job than the average traditionally published fictions in my country, at least when the original works are in English. Every single time I watch a subbed movie there will be some lines noticeably wrong.

Translators already started losing jobs due to machine translation a decade ago (e.g. DeepL), before LLMs. Remuneration going down made it more difficult to make a living as a translator already then, even if you still received offers.

Not all translations are the same. Literary translations are often works of art in and of themselves, and automating them would be missing the point entirely, like automating homework or weightlifting at the gym. I don't really know what's the state of the art, but I do buy that, on the other hand, translating toaster manuals or generic copy could soon be automatic.

"Could not connect to translation service" was apparently good enough for someone, so the bar must be extremely low.

https://www.reddit.com/r/funny/comments/3e786n/chinese_hair_...

On the other hand, a lot of people become extremely put off by the smallest sign of ai slop. And the llms have a tendency to impart their style to any text they touch.

It'll be a similar theme for all facets of work involving any language, slowly - human or code. We'll parrot about humans in the loop this and that, but I think it'll be less humans in the loop over time and I think most people will even be willing to settle for a slightly more mediocre translation or coded project. It all comes back to our dopamine addiction, where we like fast feedback. And the oligarchs like tools to suppress wages. We will be our own demise for not advocating for either UBI or job protections, instead, happily using the technology while also rolling our eyes that it could never replace us.

Is it that unfortunate? Tasks that don't require high-quality translation now don't need human labor. We should be celebrating.

The sad part is that we haven't figured out how to distribute our resources fairly to these people even thought their services aren't required as often. Instead we just take their wages and give them to the top 0.1%

Sure, just like AI enthusiasts seem to be unfamiliar with the concept of local maxima...

It’s been basically the same for 3 years now. Are you sure we’re the ones who can’t see trends?

The point of the comment was that models are improving a lot every release, so if your livelihood depends on something, you might want to check to see what the latest models are capable of before someone else (like your employer ) tells you.

The other person in the gym was right, did you you just dump it in the latest model?

she did. Did you remember to read the article?

No one assumes that AI systems won't be improved upon. What people don't assume is that progress will be infinite in every domain cheaply forever.

> LLMs make it really easy to misunderestimate the complexity

In my experience this is a real problem. Just yesterday I asked my LLM to create a piece of software that could help me build an 'ambilight-like experience' through my home assistant. It did something that seems to work as I expected, but there is a lot of theory that I just brushed past. It would be pretty easy for me to assume that I would be able to replicate this feature from scratch 'now that I understand the problem'.

Agreed. LLMs are really terrific at sounding like they know exactly what they are talking about. Fable is the best yet. Beautiful, thorough explanations with absolute certainty, which under even light scrutiny turn out to be mostly bullshit.

I still love the tool, but remain as convinced as ever that AGI does not lie at the end of this particular path.

> we will be literally inferior soon

This plague of misanthropic doom is itself pretty depressing. Why do so many people think LLMs are in any way on a path to compete with human brains? Why do you think so little of yourself? The brain is magnificent and complex in ways that we are unable to decipher anytime soon, and it does way more than an LLM. Way, way more.

The thing is that AI is not some inevitable force of nature that must just be contended with and weathered. It is an active choice by our society to develop it and it is a choice by our society how we should use it, if at all.

We would all do well to remember that and remember that each and every advancement and use case regarding AI is the result of choices by people (or the groups of people we call corporations) and are oftentimes motivated by the profit motive, not the best interest of humanity.

We could make different choices up to and including our own Butlerian Jihad where we ban all forms of AI but we could also do everything we can to prevent the worst fallout short of that.

There are only two types of problems in the universe: 1) those posed by the laws of physics 2) those posed by human choices

The problem of AI is one of the latter.

This is anecdata, but in my experience with myself and my coworkers, it is not that we believe the AI will be more accurate in software engineering, but that the answer will come faster and be more tailored to our exact problems. If I have to search SO, I have to find the answer and then tweak it to fit my codebase, but with AI tooling, the AI is already basing its answer around my code.

> there will be a point where we will not trust human input without counter check by AI

That's nonsense. There is zero reason to believe that AI (with the current techniques) will ever become reliable enough to let it do its own thing, let alone better than a human. It's been years of development and you still can't trust it to get basic facts correct, not even "well it's better than it used to be". Saying it'll replace humans in 5-10 years is a fantasy (or a prediction that people are stupid enough to fall for hype, I guess).