Warning: Some posts on this platform may contain adult material intended for mature audiences only. Viewer discretion is advised. By clicking ‘Continue’, you confirm that you are 18 years or older and consent to viewing explicit content.
I think translation is where LLM could truly shine the most. Some simpler models are literally searching for the closest meaning in the higher dimensional feature space. Translation isn’t that far off from what those models do.
I use ChatGPT to romanize song texts from Farsi squiggly lines into something readable. There are some other sites that do that, but they are all terrible and use regex replacement (I assume) and that doesn’t really work for most things since vowels in Farsi (and Arabic too) are diacritics and are often left out entirely, so you get something unreadable. ChatGPT does a fine job, but you have to make multiple, smaller requests instead of a single big one or it starts hallucinating and/or repeat passages it already romanized.
The tech is interesting, no doubt. It’s very effective as a tool to generate text nobody reads, like the marketing speak on your random startup website. It still isn’t efficient on things where what is generated actually matters.
Your example with customer service is news to me, thanks. On my end, I remember the bad experience customers had with Air Canada. We’ll see how this grows in the future.
I had a discussion last week with people saying it’ll automate software engineering, which is not a given. You say “yet”, but I’m skeptical it’ll ever work. I can see it designing UI better than a non-specialist, but the flaws in quality means I can’t trust it anywhere near my code, even though I can see a future for it as a fancy static analyzer.
People are pretending as if job replacement happens all at once, and that’s just not how it works.
A new tool that makes a job 15% more efficient will either produce 15% more goods or reduce the required labor by 15%. Some of that labor is absorbed elsewhere, but there was still a 15% reduction that happened.
Slow improvements are undoubtedly a good thing, that means we can create positions as fast as we make them obsolete. Maybe LLMs have reached their peak and we don’t have to worry about it, but it’s not a bad idea to prepare for that possibility that they continue getting better.
People really like shitting on overhyped new technologies, but I don’t think people appreciate just how big of a deal it is that a pretty basic algorithm is able to process natural language at all.
A hammer doesn’t replace a carpenter. That’s what I meant when I said that this won’t replace us: new tools are nice but they won’t automate everything. There are some jobs that have been completely replaced with advancements in technology. However, most of them have just gotten simpler and have evolved.
I do think LLMs are important, but I’m just laughing at the hype surrounding it, and all the grandiloquent claims made by tech bros.
I think a better analogy would be something like a loom: it doesn’t operate independently and still requires an operator and mechanics, but it eliminates the need for rows and rows of weavers to complete the same amount of work (and that both puts many people out of work and undercuts the labor market, which are both big problems). Judging LLM’s on a scale of total job replacement is IMHO a little ridiculous, because unless those LLM’s are fucking sentient and autonomous, they’ll never completely ‘replace’ a human roll. They will certainly make programmers/writers/translators/media producers more productive though, and that’ll put quite a few out of work, and that’s kind of a big problem.
November 2022: ChatGPT is released
April 2024 survey: 40% of translators have lost income to generative AI - The Guardian
Also of note from the podcast Hard Fork:
There’s a client you would fire… if copywriting jobs weren’t harder to come by these days as well.
Customer service impact, last October:
And this past February - potential 700 employee impact at a single company:
If you’re technical, the tech isn’t as interesting [yet]:
Overall, costs down, capabilities up (neat demos):
Hope everyone reading this keeps up their skillsets and fights for Universal Basic Income for the rest of humanity :)
I think translation is where LLM could truly shine the most. Some simpler models are literally searching for the closest meaning in the higher dimensional feature space. Translation isn’t that far off from what those models do.
Yep. They’re language models, after all. Not surprised they’re taking translation jobs.
I use ChatGPT to romanize song texts from Farsi squiggly lines into something readable. There are some other sites that do that, but they are all terrible and use regex replacement (I assume) and that doesn’t really work for most things since vowels in Farsi (and Arabic too) are diacritics and are often left out entirely, so you get something unreadable. ChatGPT does a fine job, but you have to make multiple, smaller requests instead of a single big one or it starts hallucinating and/or repeat passages it already romanized.
Definitely
The tech is interesting, no doubt. It’s very effective as a tool to generate text nobody reads, like the marketing speak on your random startup website. It still isn’t efficient on things where what is generated actually matters.
Your example with customer service is news to me, thanks. On my end, I remember the bad experience customers had with Air Canada. We’ll see how this grows in the future.
I had a discussion last week with people saying it’ll automate software engineering, which is not a given. You say “yet”, but I’m skeptical it’ll ever work. I can see it designing UI better than a non-specialist, but the flaws in quality means I can’t trust it anywhere near my code, even though I can see a future for it as a fancy static analyzer.
Where can i subscribe for daily consize ai facts?
;)
People are pretending as if job replacement happens all at once, and that’s just not how it works.
A new tool that makes a job 15% more efficient will either produce 15% more goods or reduce the required labor by 15%. Some of that labor is absorbed elsewhere, but there was still a 15% reduction that happened.
Slow improvements are undoubtedly a good thing, that means we can create positions as fast as we make them obsolete. Maybe LLMs have reached their peak and we don’t have to worry about it, but it’s not a bad idea to prepare for that possibility that they continue getting better.
People really like shitting on overhyped new technologies, but I don’t think people appreciate just how big of a deal it is that a pretty basic algorithm is able to process natural language at all.
A hammer doesn’t replace a carpenter. That’s what I meant when I said that this won’t replace us: new tools are nice but they won’t automate everything. There are some jobs that have been completely replaced with advancements in technology. However, most of them have just gotten simpler and have evolved.
I do think LLMs are important, but I’m just laughing at the hype surrounding it, and all the grandiloquent claims made by tech bros.
I think a better analogy would be something like a loom: it doesn’t operate independently and still requires an operator and mechanics, but it eliminates the need for rows and rows of weavers to complete the same amount of work (and that both puts many people out of work and undercuts the labor market, which are both big problems). Judging LLM’s on a scale of total job replacement is IMHO a little ridiculous, because unless those LLM’s are fucking sentient and autonomous, they’ll never completely ‘replace’ a human roll. They will certainly make programmers/writers/translators/media producers more productive though, and that’ll put quite a few out of work, and that’s kind of a big problem.
I agree the comparison is asinine. But I keep seeing the tech bros make it, which is what I find so funny.
Never gonna happen.
Not… with… that?… attitude?!?
:p
Guy can dream