This week’s been a pretty big one in the way of artificial intelligence (AI) advancements.
On Tuesday, OpenAI announced the launch of GPT-4, not even half a year after launching ChatGPT back in November 2022. And it’s already in use at Duolingo — read more on that down below. Plus, Google’s launching an API for its large language model (LLM), PaLM as well, an announcement that came on the very same day.
Plus, we’ve got news on a machine translation (MT) engine specializing in Finno-Ugric languages, and press releases from Acolad, Toppan Digital, and Utah Valley University.
Got news you’d like to share in next week’s Week in Review? Send it over to our editorial team here.
Duolingo’s Max Subscription Uses GPT-4 for AI-Powered Language Learning (via PC Mag)
Though it was just launched on Tuesday, GPT-4’s already being put to use: Duolingo’s using it to power new features on its AI-powered premium subscription, Duolingo Max.
Users who subscribe to Duolingo Max — for $30 a month — can access two new features, “Explain My Answer” and “Roleplay,” both of which employ GPT-4 technology to allow users to more actively engage with their target language. Currently, this subscription is only available to English speakers learning French or Spanish, but the company hopes to launch more AI-powered courses soon.
Tartu University’s machine translation engine now translates 23 Finno-Ugric languages (via Estonian World)
Finno-Ugric languages are spread across a wide territory and are the native tongue of several million native speakers (mostly speakers of Hungarian and Finnish).
Yet many of these speakers have not been able to take advantage of advancements in machine translation (MT) to the same extent that speakers of languages like English and Spanish have. Researchers at Tartu University in Estonia recently added 17 Finno-Ugric languages to their MT engine, many of which are now available on a public MT tool for the first time in their history.
Google opens up its AI language model PaLM to challenge OpenAI and GPT-3 (via The Verge)
It seems like every company in big tech is working on a LLM these days.
On the same day that OpenAI announced the launch of GPT-4, Google also announced that it’s launching an API for one of its LLMs, PaLM. Unlike Google’s other LLM, LaMDA, PaLM is not trained mainly on dialogue, so some experts say PaLM could be bigger competition for OpenAI’s GPT series.
A new and better way to create word lists (via Science Daily)
If you have to generate word lists for your research projects, you may find LEXpander to be an interesting method for creating those word lists.
Researchers at the Complexity Science Hub in Vienna recently published a paper detailing the development of LEXpander, their new algorithm for expanding word lists. According to the researchers, LEXpander performs significantly better than other state-of-the-art algorithms on word list expansion tasks, and is also capable of generating word lists in English and German.
Press Releases and Event Recaps
- Acolad appoints Bertrand Gstalder as new CEO
- New course “Interpreting in Educational Settings” now available from the Academy of Interpretation
- UVU Department of Languages & Cultures to host Language Services Industry Day 2023
- Toppan Digital Language receives “fastest growth” award in the Nimdzi 100 Ranking of language service providers
- GLOBO reaches single-digit connect times for telephone interpreting services, announces financially-backed guarantee
- Acolad announces Arnaud Daix as new Executive Vice President for North America
- Reflecting AI-driven changes in the localization industry, TOFT International adapts services