Notes · 2025
Short posts from LinkedIn and X.
-
I was dreaming of this when we started creating the Learning via Subtitling platform 20 years ago. Isolating dialogue was impossible if you didn't have access to the audio tracks.
-
Honoured to deliver the keynote at ELIA - European Language Industry Association 10th Focus on Project Management in Athens this December.
This milestone edition invites us to reflect on how we can stay sharp and ethically grounded in an AI-driven world.
Looking forward to joining this agora of project and production managers for three days of exchange and learning.
-
What if, instead of glowing brains to represent AI, people used photos of data centers?
Google’s data center in Santiago, Chile. Marcos Zegers for The New York Times.
-
Ποιον όρο προτιμάτε; Λαπάς ή σαβούρα; Ή κάτι άλλο;
Επίσης: ενσκάτωση ή σκατοποίηση;(αναρωτιέμαι τι κάνει η αυτόματη μετάφραση σ' αυτήν την ανάρτηση).
-
I’ll be celebrating International Translation Day on the radio 🎉 Actually, more like International Translation Night. The good thing about radio is you can talk in your pjs. I don’t know what Nefeli Lygerou will ask me, but I’ll be talking about what I love most in my work as a subtitler.
Tune in to hear me rambling in Greek - or listen to the replay tomorrow.
-
AI is that magical stuff that has converted computer scientists into experts in ethics and philosophy.
-
Tasks are not jobs. The fact that a human or a machine can perform a task successfully does not mean that they can do the job.
#spoiler AI Localization Think Tank
-
“Brevity is the soul of wit,” said Shakespeare.
#Subtitlers work with brevity and conciseness for a living.
Which makes them true guardians of wit.
-
Turns out the tech world is also able to see it: "words aren't just words, they carry trust, context, and culture".
And if you thought you hired Shakespeare but got Clippy with ChatGPT installed, and your entire brand voice now sounds like an airport pamphlet - translators can fix it for you.
If you hired back your content people, don't destroy their work in other languages.
-
What's the deal with films these days? Stunning locations, gorgeous actors, some of them actually good too, obviously lots of money spent. But the script? It sucks. Hollow words, weak plots.
Dear Hollywood, could you pls become more attractive for scriptwriters?
Yours,
A subtitler -
"It'a not about X, it's about Y. But here's the thing."
Did everyone write like this before? It's the new "navigating the realms". And it's really boring.
-
Look forward to a transparency label integrated in tools so that it can be easily used. For translation too.
-
I've always thought that white robots and blue glowing brains are misleading representations of AI*. Better Images of AI suggest they are potentially harmful.
Let's do better (looking at you, dear journalists).*what we currently understand as artificial intelligence.
-
How much do you hate repeating a task? I hate it so much that I prefer spending the same amount of time (or even more) to automate it. At least I learn something and it's more rewarding. But it's also a trap, timewise.
-
Are thermostats agents? They choose to turn on or off depending on the environment.
"Agency" is a fascinating concept and we can spend hours discussing. Thinking of automation as a continuum, like Marina suggests, is quite helpful. At least we can agree to disagree.
Shall we keep "agency" for Marina's level 5 then?
-
Been flooded with Veo3-made videos, and impressed by lipsync, accents, voice delivery. Thinking in what ways it changes the conversation on #AIdubbing
-
Anybody remembers #promptengineers? Didn't last long. No #vibecoders mentioned in this article either.
Lots of #integrators, #designers - even #AIplumbers, cleaning up the pipes when agentic workflows don't flow.
A.I. Might Take Your Job. Here Are 22 New Ones It Could Give You.
-
Justin Beaudin has this rare ability to see things with clarity and openness, zooming out to grasp the bigger picture, grounded in the present but also focused on the future. He works with AI but in a truly human-centric way.
I totally recommend listening to him in this podcast, so grab your headphones and head out for a walk, a run, or whatever gets you moving.
-
"Always do your best, and submit something to be proud of!"
That's what the long email with instructions for translators ends with. The rest is "Follow the guidelines, meet the deadline, here's the [unsustainable] fee". A pittance really.Are you doing your best, dear LSP? Maybe you have no other way: you are squeezed, so you squeeze further down. But are you proud?
-
In light of tomorrow's event on #MediaLocalization by Women in Localization - Greece Chapter I thought I'd recheck Heygen and see if there are any developments in the last two years.
I used the video I made for RAIN Film Festival.It's eery because I don't speak Spanish with a slight Latin American accent.
-
Beyond the title (no serious person does that anyway), there are interesting insights re prompt improvement. They also provide alternative wordings, such as Intermediate Token Generation instead of "thinking" or "reasoning", I wish they would catch, but not optimistic, as "factuality deviation error" didn't catch either, we still use hallucination.
-
Spending energy to find out how much energy I spend.
With this tool that tells you how much energy your AI chatbot uses.
-
How often do we get a #NewYorkTimes newsletter featuring #translation?
Cherishing the moment, and thinking that prizes and #awards make headlines. It's a way to reach the wider public.
#DeepaBhasthi #HeartLamp #Kannada
-
This essay by Luciano Floridi has been in my reading list for a few days, but I stopped at one word, something in me crumpled. WrAIting?
I wonder if it will stick, like that "e" added everywhere in the 2000s (e-mail, e-banking). For the moment, I hope not, or any use that converts "i" to an "AI".
-
Being coached is a luxury we all need to experience at some point in life. And if that coach is Magda Jagucka Miller you get a masterclass in clear communication & active listening. You're not just guided, but challenged to think clearly and re-focus, so that you can design your course of action.
Thank you, Magda!
-
Dear heads of localization departments of large companies, this one's for you.
You probably know all about how to prove your/our value in the age of AI, but this article sums up some of the most important points.
Expertly written by members of the AI Localization Think Tank
-
Ένα χρήσιμο λεξικό με όρους και επεξηγήσεις περί τεχνητής νοημοσύνης στα ελληνικά.
(Θα μου είχε γλιτώσει πολύ χρόνο πριν από δύο χρόνια όταν έψαχνα τον σωστό ελληνικό όρο για την Generative AI. Αλλά ήταν μια χρήσιμη άσκηση, παρόλα αυτά.
«Μικρό Λεξικό Τεχνητής Νοημοσύνης» – Εγχειρίδιο του Γιάννη Φαρσάρη
-
Knowing AI well enough to spot human bullshit is an underrated skill.
-
Have you noticed that kids and teens don't watch long-form content any more? Will it change when they grow up? Are millenials the last generation to read books and watch films?
-
Love this - though it’s sad to think there may be no soil left for planting scented flowers or water to keep them alive.
Too many discussions with frustrated colleagues lately, especially in #subtitling.
#translation #PDOhumanity #humanmade
-
"A non-human machine" is an expression actually used in a court decision.
I suppose it's time for me to accept the "human translator" now.
-
Just started the Introduction to #AISpeechEditing course by Voiseed and OOONA.
I was so excited about it that, despite being pressed for time this week, I just wanted to take a quick peek. And I ended up finishing the first unit already 🌟I’ll have to be more disciplined and continue next week.
-
Today I read a piece where the author used "the essence of my thinking" instead of "TL;DR" and I found it beautiful.
People use it casually in internet-driven discussions, and fair enough. But what began as a snarky brush-off, "too long; didn’t read" has settled into a less dignified way of saying here’s the summary. It gets the job done, but...
-
I need an AI to tell me which AI to pick for what I want to do.
Next stop?
I asked an AI, and this was the answer:"Next stop: An AI that will tell me what I want, send my request to the right AI, and then bill me for a premium subscription to overthinking-as-a-service."
#AIinception, #advanced_promptengineering
-
#OpenAI always sold itself as a global initiative, not as a US-centric company. "Build AI that benefits all of humanity", focus on openness and collaboration. But the clash with #DeepSeek tells a different story. There is an open shift to a more nationalistic approach, an emphasis on protecting its technology from foreign entities and a commitment to working with the US government.
-
Shall we keep in mind when using the term #ResponsibleAI that #AI cannot actually be responsible? That it can only be *used* responsibly?
Sure, it’s a shortcut people use to emphasize the collective responsibility for ethical use, minimizing harm, maximizing fairness, transparency, accountability.
But it’s worth reminding ourselves now and then that the responsibility lies with us. -
Doing some storage cleaning, I was reminded of the EST congress in Granada, 1998, with Jorge Díaz-Cintas, Eliana Franco (obviously when we were 10 years old) yves gambier and Henrik Gottlieb.
