Alexander Kustov

Thanks, this is helpful. I appreciate you trying to help me understand folks like you and prof. Bender.

Proud to accept this award, thanks. Seriously though, this isn't 2020 anymore—you can't just call people names to shut them up. I tried to have a good faith dialogue about whether LLMs can translate text. The response was name-calling & "I see no beneficial use cases for synthetic text." In 2026. OK

Today's mansplainer award goes to the prof in (checks notes) a School of Global Affairs, who seems to believe that he knows better than I do about how language technology works.

What about agentic tools like Claude Code or Codex? If my mom uses one to help identify and order prescription medicine from her doctor, is that also not beneficial? What if a deaf person uses AI to generate real-time captions? Zero beneficial use cases is a wild position.

As I also mention there, if you haven't noticed, the post was AI-generated but based on my earlier social media writing and ideas. You can think of it as something in-between a helpful and expensive professional editor or a ghost-writer if you want to keep being mean :)

Thank you! I really want to hear @emilymbender.bsky.social say "The LLM technology is a remarkable achievement with all kinds of impressive use cases like translation [even though I disagree with how it's done]." That's it. No gotcha of any kind. Just literally concede a tiny point without any smug.

I think my writing skills are fine, thanks. I did published quite a few things before LLMs were even around, but I appreciate your concern about my ability to discover new ideas, truly.

So, uh. In disability studies we call this a "disability dongle" or a techno-utopian solution to a problem that appeals to the able-bodied imagination rather than the actuality of disabled life itself.

Deploying a LLM to translate a legal notice for an asylum seeker functions similarly.

Thanks, I appreciate it. I meant most popular languages, of course, with a focus on immigrant-receiving countries where my work matters. I left the original English titles for consistency with other languages, but you're right that it may be messing up word order. I'll try to fix it now.

Thank you, I appreciate it. I didn't know that there was a norm against quote-posting here. Now I know it even if I disagree with it.

Besides, you can write with AI in your voice—you can train it on your writing and specify your idiosyncratic preferences. Not everyone has access to a professional editor.

This is certainly a good use of these tools, and probably what I use it for most of the time as well. I do think though that, for me as a non-native English speaker, it can help improve and tighten some of my scattered thoughts.

Study suggests telling people how burdensome or restrictive migration controls are does more to shift their views on migration than trying to persuade them of the benefits of migration

In the meantime, I just used these stochastic parrots to redesign my website, publish it, and translate all my writing into most world languages in less than one day.

Check it out and decide for yourself whether the quality and ethics of it is good or not: alexanderkustov.org

What's frustrating here is the conflation of two very different things: empirically verifiable facts about what LLMs can do well (like translation, which is at this point undeniable) and normative disagreements about whether using them is ethical. You can debate the second without denying the first.

I actually agree with all three. But here's one genuine question for you and your followers: can LLMs also help translate simple texts well?

If so, just to make things concrete, would it be wrong for, e.g., an asylum seeker to use one to translate a legal notice into a language they understand?

Academics Need to Wake Up on AI Ten theses for folks who haven't noticed the ground shifting under their feet

I'm referring to the prevalent idea that only artisanal, hand-crafted arguments are legitimate or that these artisanal arguments are more legitimate than the same arguments produced by AI. I wrote more about here:

www.popularbydesign.org/p/academics-...

Still, even though I disagree with existing scholarly policies against AI use, I will of course abide by them fully. To be clear, this is about how I approach my own work.

Using AI responsibly also means respecting confidentiality and consent when it comes to other people's data

The problem with AI in scholarly writing is not using it per se but using it carelessly without doing the work and verifying the claims. But this is true for any bad writing and argumentation, whether AI-assisted or not.

Research is about uncovering truth, not self-actualization.

I see many folks are pledging not to use AI in their writing.

I pledge the opposite: I will use the latest LLMs, and for that matter any other available tool, to best improve my research or the way I communicate it.

That way, if my name is on it, you can be sure it reflects my own best judgment.

Thanks, yeah, that's the plan! For folks to be able to discover my stuff more easily when they search in non-English languages.

If there's enough demand, I can create a Claude skill for this. But it's all easy to learn by yourself: just ask CC to do it for you and answer its Qs about your preferences.

Incidentally, I wish Substack would allow for a more seamless translation on their own website.

alexanderkustov.org

If you're wondering about my own setup, I did it with Claude Code Opus 4.6 using a committee-of-translators approach and personally verifying the languages I speak.

Still, mistakes are possible. If you spot errors in your language or want to suggest adding another, let me know.

As educators, we want our research to be as accessible as possible, and now we can actually make that happen given our limited resources.

I still see many junior profs without a personal website. You can now build one in a day and translate everything automatically in another.

I've been getting more requests from international folks to adapt my writing, so... I decided to translate my entire academic website and newsletter essays into most world languages using Claude Code myself as a side quest.

Let me know if anything needs fixing, but you should probably do it too!

Yes. Workers of the world, unite!

Skills on the Move: Linking Labour Mobility and Vocational Training Technical and vocational education and training (TVET) is often touted as an important tool in low- and middle-income countries struggling with high levels of unemployment and large skills gaps. As a ...

Imagine visionary collaboration to train migrant workers with skills critical to the energy transition.

Investing in migrants’ human capital as a strategic resource.

The brilliant @helen-dempster.bsky.social @cgdev.org and @irexintl.bsky.social are not just imagining it. They’re building it —>

I feel for you—publishing a painstakingly researched but outdated book. I don't know you or want to question your integrity.

But last time I checked, posting takes on a personal blog, whether AI-assisted or not, is neither illegal nor unethical. Nobody is forcing you to agree with or even read them

Prof. @emilymbender.bsky.social, I appreciate you sharing your opinion on my integrity, which you're free to do even if it's not informed.

But you're tenured too, so it's OK to admit you weren't right. You can say LLMs were trash before but now they can actually be useful to folks who use them.

As someone who has probably more credibility among folks much further to the left from me, you bringing up is very much appreciated.