Martin Ruskov @mapto@qoto.org

Today at #CHR2025, I will be presenting our work on the evaluation of the historical adequacy of masked language models (MLMs) for #Latin. There are several models like this, and they represent the current state of the art for a number of downstream tasks, like semantic change and text reuse detection. However, a historical researcher, philologist or else would want to be sure that such models really represent the historical period of interest. For example, it would be an embarrasing hallucination if St. Augustine showed up in the context of the Roman senate.

Our evaluation confirms a known problem: LLMs and masked models in particular are trained on corpora without attention to historical periods. Unlike other research we've done on Early Modern English, this problem leads to models being barely distinguishable when it comes to their ability to generate based on a historical period. Even though history is a case where it is most obvious when models go wrong, this type of contamination is a known problem for LLM training overall, think of different legal jurisdictions using the same language, dialects in programming languages, etc.

This research was generously supported by AgileLab.

The full paper is available at:
https://anthology.ach.org/volumes/vol0003/the-latin-language-evolved-over-time-masked-models/

"The results speak for themselves. Today, Uruguay produces nearly 99% of its electricity from renewable sources, with only a small fraction—roughly 1%–3%—coming from flexible thermal plants, such as those powered by natural gas. They are used only when hydroelectric power cannot fully cover periods when wind and solar energy are low. The energy mix is diverse: while hydropower accounts for 45%, wind can contribute up to 35% of total electricity, and biomass—once considered a waste problem—now makes up 15%. Solar fills the gaps.

The economic impact has been profound. The total cost of electricity production decreased by roughly half compared to fossil-fuel alternatives, and the country attracted $6 billion in renewable energy investments over a five-year period—equivalent to 12% of its GDP. About 50,000 new jobs were created in construction, engineering, and operations, roughly 3% of the labor force. Even more striking, Uruguay is no longer subject to the wild swings of global fossil fuel markets.

This transformation was not just technical; it was also regulatory and structural. Uruguay moved to long-term capacity markets, providing investors and utilities with predictability while removing the bias that favored fossil fuels. The government’s adaptive approach, maintained through five administrations, ensured consistency. Instead of making climate the primary focus, policymakers prioritized cost, reliability, and economic benefits; emissions reductions were a valuable bonus."
https://www.forbes.com/sites/kensilverstein/2025/10/19/uruguays-renewable-charge-a-small-nation-a-big-lesson-for-the-world/

Maybe an year ago, I was listening to an interview with Albanese where she spoke how powerless she was to stop genocide in Gaza. I thought that as an UN rapporteur, she only had to try and would change much. Later she tried, and got smashed by US politics in a blatant eradication of any civic common sense.

I guess I feel that in some way it is my fault. People like me who have encountered her have nudged her, and now she pays the price. I still feel it is a small price one in her position could pay for trying to help millions of innocent victims, but it is not my judgement to make.

Francesca Albanese is an everyday hero!
https://www.aljazeera.com/video/newsfeed/2025/12/6/francesca-albanese-tells-al-jazeera-us-sanctions-have-made-her-non-person

I learned about Repair Cafe by the news its funding is discontinued. But more cities need to support this type of events. In a world of planned obsolescence, the environmental bill is being paid by communal waste collection services. A targeted analysis of which repairs could be most cost-effective for waste management could help justify funding of such initiatives with volunteer-based repairs.
https://spec.bc.ca/waste/repair-cafe/

From distant times, probably when Disney was still a person and not a moneymaking machine

https://youtube.com/watch?v=8BqnN72OlqA&pp=ygUPbWF0aCBtYWdpYyBsYW5k

Very interesting reflections on how AI slop is overtaking influencers. It seems to me that most shortcomings of slop mentioned here are going to get overcome (under the assumption that money poured continues to be practically unlimited). Probably the one that is not mentioned here but is going to stay, is the lack of integrity across video cuts.
https://www.theverge.com/entertainment/839494/ai-literacy-tiktok

This article suggests that research in 3D rendering contributed to machine- assisted assassinations of children in Gaza.

It is a topic very close to my heart. When I was starting my academic career at the dept. of Digital Storytelling of ZGDV in Darmstadt, Germany, I took a lot of inspiration by colleagues from the Institute of Creative Technologies at the University of Southern California. Their work was funded by DARPA and a lot of it was about military training. Let alone the fact that training military personnel through storytelling is not about assassinations at all, we've had conversations about this. In one we were told something along the lines of: You European have [what is now Horizon]. In the US our federal funding is DARPA [coming from the US DoD].

This incident is one of several that persuaded me to a widespread narrative in information security. It states that making research open access is a way to contain its destructive use. I still tend to believe so. Research at Pixar and the ICT we all publicly available at their websites. Would you agree that tracing the blame down to people that invented technologies that have subsequently emerged as dual use is misleading? I still think that assassinations of children and civilians in Gaza or elsewhere are a direct consequence of a chain of political, commercial and military decisions. This is where responsibility needs to be investigated.

https://english.elpais.com/technology/2025-12-02/how-pixars-technology-helped-develop-more-lethal-military-drones.html

The Chinese side of AI safety is glimmer than Musk's dreams
https://www.aspi.org.au/report/the-partys-ai-how-chinas-new-ai-systems-are-reshaping-human-rights/

UN report: Israel silently tolerates torture and murder in its prisons
https://www.theguardian.com/world/2025/nov/29/israel-has-de-facto-state-policy-of-organised-torture-says-un-reportses