X . com utzi eta BlueSky-n kontu bat irekitzea tabakoa utzi eta bapeatzen hastearen modukoa da: modernoagoa dirudizu, baina adikzio-arrisku antzekoa izango duzu, eta atzetik dagoen negozio-eredua berdina da. ICE ongietorria den plataforma honetan ez naiz eroso sentitzen, banoa berriro Mastodon.eus.

As Morocco, Mauritania and Senegal clamp down on unauthorized migration, boats are increasingly leaving from much further south, in Guinea, to reach the Canary Islands.

My latest on Substack -- a write-up of the talk I gave at NeurIPS in December. aiguide.substack.com/p/on-evaluat...

This analysis of papers from 1980 to 2025 goes beyond LLMs, but notes one challenge that arises when research questions rely on the same tools and datasets: “the adoption of AI seems to induce authors to converge on the same solutions to known problems rather than create new ones.”

Some important new government policy that Native Americans might be interested in:

We updated the plots we use to measure the open model ecosystem at interconnects, to guide The ATOM Project, and to understand what's happening. We have ~8 plots to summarize what's happening. First, the high level picture showing China's growing adoption lead.

Israel is a rogue apartheid state committing genocide. Palestine WILL be free. Happy New Year! 🇵🇸🇵🇸🇵🇸

Sorry not sorry 😘

I'd like to propose the following norm for peer review of papers. If a paper shows clear signs of LLM-generated errors that were not detected by the author, the paper should be immediately rejected. My reasoning: 1/ #ResearchIntegrity

Update: the ACM DL is being updated so that author-provided abstracts are the default view again, and AI summaries are more explicitly labeled with "AI-Summary". These changes should be visible worldwide by sometime tomorrow.

Shame on the ACM. They should know better. In fact, here's a fun exercise to the reader: How many points from the ACM's own code of ethics does this move violate? www.acm.org/binaries/con...

Just so I'm clear on this, computer memory has tripled in price because a bunch of it that hasn't been produced yet has been ordered to populate GPUs that aren't installed in data centers that aren't built yet in order to service a demand that doesn't exist to make profits that don't happen.

I'm working on a piece about peer review and conference publication in AI/ML/NLP/CV, and I need YOUR help to understand how the community uses conferences and peer review in paper discovery. Your four short multiple choice answers will help a lot :)

Merriam-Webster’s human editors have chosen ‘slop’ as the 2025 Word of the Year.

Worth a read: Tim Dettmers (CMU, AI2) on AGI and ASI. timdettmers.com/2025/12/10/w... My skepticism has also relied on the physics of hardware. Tim goes into a lot of detail.

Euskaldun izateaz harro ✊🏻 www.theguardian.com/football/202...

Uda honetan @zubidasoa.bsky.social proiektua @mtsummit2025.bsky.social konferentzian aurkeztu nuen. Harrera ona izan zuen, bertaratu zirenek interes handia adierazi zutelarik. Zuen interesekoa ere izatekotan, hemen uzten dizuet aurkeztutako artikulua (ingelesez): aclanthology.org/2025.mtsummi...

The MT Summit 2025 program is now live! mtsummit2025.unige.ch/programme.ht... 📅 Don't forget to register before April 30 to benefit from early-bird rates!

ZuBidasoa project will be presented @mtsummit2025.bsky.social More info at @zubidasoa.bsky.social

"The conclusion is clear: Israel is committing these acts with the intent to destroy the Palestinians in Gaza. Israel is committing genocide. We also found that not only is the genocide in Gaza the most documented in history, but the EU and many of its member states are failing to prevent it."

ZuBidasoa doktoretza-ondoko ikerketa proiektu bat da, migratzaileen beharretara egokituko diren hizkuntza teknologiak modu parte hartzailean garatzea helburu duena. Proiektua EHUko HiTZ zentroaren eta CNRSko IKER (UMR5478) zentroaren artean garatzen ari da, 3-4 urteko epean.

As arXiv links accumulate, I'm enjoying my read of "The Routledge Handbook of Translation and Migration". Truly interdisciplinar approach, centered on people and language, including two chapters about MT. 👇🏻 www.routledge.com/The-Routledg...

I have opened an account for promoting my current research project. If you are interested in migration and language technologies, follow @zubidasoa.bsky.social (Most of the messages will be published in Basque. If needed, in-house MT works fine enough for translating into high resource languages.)

TESSFIT TEMZIDE YAYA KARAMOKO ABDOULAYE KOULIBALY MOHAMED KEMAL FAYÇAL HAMADOUCHE AHMED BELHIREDJ SOHAIBOU BILLA IBRAHIM DIALLO ABDERRAMAN BAH Iturria: Bidasoa 2018-2023 Irudia: Mattin Sormena

It may not be much longer until we know of an LLM being directly involved in a kill order by a military or government www.972mag.com/israeli-inte...

AI real talk. We (humanity) are moving full speed ahead at building AI agents for war that can create a runaway missile crisis of mutually assured destruction globally. Is the option of not allowing AI agents to deploy missiles already off the table, or is that still up for discussion?

While we look forward to a sunny Geneva, why wait to join the conversation? We’ve created a starter pack for our #GITT2025 friends! 🕵️ Follow researchers working on gender bias in MT 💬 Stay up to date and dive into the discussion! All info at sites.google.com/tilburgunive...

🎉 Mystery AI Hype Theater 3000 has reached Ep 50 🎉 Dr. @tamigraph.bsky.social joins @alexhanna.bsky.social & me to look into tech bros’ unhinged ideas about energy & the environment & the techno-fascism of it all. www.buzzsprout.com/2126417/epis... Thx to @whatulysses.bsky.social for production!

A preprint of my book Natural Language Generation is now available on Arxiv (arxiv.org/abs/2502.14437). Use the PDF, the HTML version is incomplete. For more info on my book see ehudreiter.com/book/ . Note it is about NLG concepts, its not about the latest LLM tech

📢 Based on the recommendation from peer review committee, ARR switches from 8 to 10-week cycles. This will allow for a saner cycle schedule. See the full report here: www.aclweb.org/adminwiki/im... The ARR website is now updated with the new submission dates: aclrollingreview.org/dates

The International Energy Agency just increased its prior growth predictions for energy demand. Data centers are one of the major drivers. Here's why that's significant: www.theverge.com/news/614356/...

Bisita izan dugu! Eusko Jaurlaritzako Hizkuntza Politikarako Sailburuordetzako talde zabalarekin, Aitor Aldasoro sailburuordea, Josune Irabien eta Sonia Rodriguez barne, arratsalde ederra pasa genuen zentroa erakusten eta Ikergaitu proiektuko azken emaitzak erakusten https://www.hitz.eus/iker-gaitu/

Error Span Annotation for machine translation. Very high-level, very quick. See you at NAACL! youtu.be/995GDufPBw8 arxiv.org/pdf/2406.12419

The Indigenous People of North America have decided to go back to calling the Gulf of Mexico, “Chalchiuhtlicueyecatl”. It is what the Nahuatl People originally called it as it is the domain of the goddess Chalchiuhtlicue, who is assoc w/water bodies, including the Gulf of Mexico.

Gogoratu otsailaren 17rarte aukera duzuela euskarazko txatbotak ebaluatu eta ikerkuntzan laguntzeko: ebaluatoia.hitz.eus Sartu eta parte hartu, erraza eta dibertigarria izateaz gain, sariak ere badaude!

ICE is Gaming Google Search U.S. Immigration is Gaming Google to Create a Mirage of Mass Deportations 13,000 archived ICE press releases timestamped to 24 January 2025. "Every article was updated on the 24th ... causing the Google SEO to...rank it higher." Brilliant work @darakerr.bsky.social!

📢 NEW: The US government permitted Nevada’s Thacker Pass lithium mine without prior consent from even a single Tribe impacted, violating indigenous peoples’ human rights. Our HRW + ACLU report says, out today. 🧵 1/4

this is really bad considering Cohere is considered one of the good, responsible AI companies around. shows there really is no good AI company like there is no good millionaire

Ikerkuntza galdera ireki bat dago mundu mailan: Euskara bezalako hizkuntza "txiki" baterako txatbotak eraikitzeko zein da teknika egokiena? Horretarako euskaldunon laguntza behar dugu, bi txatbot anonimorekin aritu eta "hoberena" zein den bozkatu. Ezetz 20 mila bozka lortu!

Euskarazko txatbot publikoak hobetzeko zure laguntza behar dugu! https://ebaluatoia.hitz.eus Sartu, galdetu eta erantzun hoberena aukeratu. Erraza eta dibertigarria, lehiaketa eta sariak daude!

I have found this podcast quite useful for finding other researchers working on topics related to language and migration. Go check it out!

Recent developments show that several independent open source initiatives are matching the best close models. This is great news for open, reproducible science, and also for smaller languages like Basque, giving a great boost to our strategy @hitz-zentroa.bsky.social

Very good (technical) explainer answering "How has DeepSeek improved the Transformer architecture?". Aimed at readers already familiar with Transformers. epoch.ai/gradient-upd...

What are the possibilities and limitations of Large Language Models like #ChatGPT? How will they impact society? And what about their biases? Prof Malvina Nissim talks about it in the latest In Science podcast ep. with hosts @doctinak.bsky.social and @wimbrons.nl 🎧 👇 www.rug.nl/about-ug/lat...

To set things clear about DeepSeek + sustainability: it (seems that) training is much shorter/cheaper/more efficient than traditional LLMs, but inference is longer/more expensive/less efficient. So it's worth thinking about the trade-offs that this has in terms of energy and emissions [...]

Distillation is violation of Terms of Service. That is not Intellectual Property theft. It’s up to the company to enforce TOS, not the government. Outputs of AI stems are not copyrightable either. Oh, and OpenAI might be violating copyright anyway.