(6/6) Many thanks to Sara Robertson and the Institute of Translation and Interpreting for granting me the space to offer a general overview of this urgent issue.
16.03.2025 14:03
👍 0
🔁 0
💬 0
📌 0
(6/6) Many thanks to Sara Robertson and the Institute of Translation and Interpreting for granting me the space to offer a general overview of this urgent issue.
(5/6) …who stand the most to gain by pushing for the implementation of AI language technologies into anything and everything. My hope is that translators and other so-called digital workers can lead the way in promoting a vision of technology that rejects exploitation of workers and the planet.
(4/6) What's ultimately needed is a reconfiguration of the structures of incentives and ownership that drive the ceaseless expansion of AI infrastructures without regard for their ecological costs. This transformation would necessarily challenge the big tech corporations (including many LSPs)…
(3/6) But simply encouraging developers to make models "leaner" and more resource-efficient is a false solution: AI's increasing adoption across sectors and companies seems to be overshadowing all improvements to model efficiency.
(2/6) …it’s admittedly difficult to parse individuals' ethical responsibility for these environmental harms. Obviously, there are numerous practical constraints that would make it hard for translators to abstain from using these technologies altogether.
(1/6) There's growing alarm over the substantial environmental harms of the sprawling infrastructures and global supply chains that make AI technologies possible. For a profession/industry like translation, which is now largely intertwined with these technologies…
www.iti.org.uk/resource/the...
Sasha Luccioni I believe is the one who did the earliest and most realistic estimate of the impact of LLMs. She assembled an excellent resource here - huggingface.co/blog/sasha/a...
As the translation industry continues raving about LLMs’ translation capabilities, it’s worth drawing attention to Luccioni’s recent study showing that such generative AI models can use up to 30 times more energy than task-specific models (like traditional NMT) to complete the same task!
🚨Call for papers: "Second Workshop on Creative-text Translation and Technology", co-located with the MT Summit 2025
👇
ctt2025.ccl.kuleuven.be/calls
Follow us for updates: @ctt2025.bsky.social
Organized with @bramvanroy.bsky.social @anaguerberof.bsky.social Lieve Macken, Damien Hansen & Paola Ruffo
