The Week in Review — Feb. 24, 2023

If you’ve been following MultiLingual for a while now, you may remember our “Weekly Shorts” series from a while back, where we rounded up a week’s worth of language and language industry news every Friday. Though we ended the series in the summer of 2021, we’re pleased to bring back a revamped version today: “The Week in Review.”

For our first installment, we’re bringing you a couple of thoughtful pieces on large language models (LLMs) — first, we’ve got a look at the so-called “good at language → good at thought fallacy,” then later, a look at ChatGPT’s machine translation (MT) capabilities. Plus, a look at Phrase’s latest wave of releases and a recent report on Indigenous language revitalization in British Columbia.

Interested in being featured in future installments of “The Week in Review?” Feel free to submit news updates to our editorial team here.

To understand language models, we must separate “language” from “thought” (via TechTalks)

You may have rolled your eyes last year when an employee at Google claimed that the tech company’s LLM had achieved sentience. For good reason, too: even if an LLM can produce highly coherent, human-like speech, that doesn’t mean it can think or feel like humans can.

Researchers at the Massachusetts Institute of Technology and the University of Texas at Austin spoke to TechTalks earlier this week about their work to debunk what they call the “good at language → good at thought fallacy.”

B.C. First Nations Languages Report Shows Increase in Language Learners, Needs Long-term Support (via the First Peoples’ Cultural Council)

It looks like efforts to revitalize Indigenous languages in British Columbia are paying off.

The First Peoples’ Cultural Council announced the release of the fourth edition of its “Report on the Status of B.C. First Nations Languages” earlier this week. Using data from 2022, the council found that there are about 3,000 more learners of First Nations languages and more than triple the number of language nests for preschool children than there were in the last report four years ago.

Nimdzi Insights Shares its 2023 Nimdzi 100 (via Nimdzi Insights)

Less than a week after revealing a sneak peek at the top six spots on the company’s flagship ranking of the biggest language service providers (LSPs) in the industry, Nimdzi Insights shared all 100 LSPs to make the ranking. Perhaps unsurprisingly, TransPerfect remained at the top of the list this year, after becoming the first (and, to this date, the only) LSP to break the $1 billion barrier back in 2021.

ChatGPT’s Translation Performance and What It Tells Us About the Future of Machine Translation (via Lionbridge)

When ChatGPT first came onto the scene, a lot of people decided to test out its translation capabilities. While researchers at Tencent maintain that state-of-the-art MT tools still fare better on MT tasks than ChatGPT, its translation capabilities are also not something to be brushed aside.

Last Friday, the folks at Lionbridge took a closer look at ChatGPT’s potential to shake things up in the field of MT.

Global Language Advocacy Day 2023 (via the Coalition for Language Rights)

On Tuesday, folks all across the globe celebrated the 24th International Mother Language Day.

The following day marked the second annual observance of Global Language Advocacy Day, an event organized by the Coalition for Language Rights. This year’s message was “Language Rights Save Lives,” and featured a handful of panels and talks on all the ways language access measures can save people’s lives in times of crisis.