Duolingo Turns to AI Translation to Expand Its Most Popular Courses to All 28 User Languages

AI is bringing a massive boost in productivity to Duolingo, maker of the most popular app for learning languages.

What’s new: Duolingo used generative AI to produce 148 courses, more than doubling its previous catalog. The technology enabled the company to offer some of its most popular courses — Spanish, French, German, Italian, Japanese, Korean, and Mandarin — in 28 languages. Initially, the company is using AI to produce courses aimed at beginners, with more advanced levels to come.

How it works: Duolingo’s AI-assisted approach to building language courses quickly turns a single course into many. The new approach revved its pace from building 100 courses over 12 years to producing many more than that in less than a year.

Duolingo starts by building a base course and uses AI to translate it into numerous languages. For example, it can adapt a course that enables English speakers to learn French into a course for Mandarin speakers.
The new process gives the company more flexibility in allocating resources, Duolingo’s head of AI Klinton Bicknell told Bloomberg. Previously, the company could dedicate a team to either creating new high-demand courses or updating an existing course. Now it can do both.
The quicker pace will enable the company to meet rising demand for instruction in Asian languages such as Japanese, Korean, and Mandarin.

Behind the scenes: AI is at the heart of Duolingo’s expansion into other areas beyond language learning.

Duolingo has used OpenAI models to build curricula since 2023. However, it is evaluating models from Anthropic and Google as well as open options.
Following one test, Duolingo concluded that Anthropic’s Claude was “much better” at generating certain types of math content for the company’s relatively new math curriculum, according to Bicknell.
The company’s embrace of AI drew criticism last week after CEO Luis von Ahn recently posted on LinkedIn that it would stop hiring contractors to do work that could be automated and increase staffing only in areas that couldn’t be automated. Since then, Duolingo has noted that it plans to hire more engineers and AI researchers, and employees will generate data used to train AI instead of performing quality reviews and other jobs that AI can do faster.

Why it matters: Companies in nearly every industry face pressure to produce more with less amid rising competition. AI can help to accomplish that while potentially improving product quality, and Duolingo has ample reason to move aggressively in this direction. The startup Speak, which offers a voice-based approach to learning languages, is growing rapidly, and Google just launched Little Language Lessons that show how an AI-first product could be used as a language teacher and conversational partner.

We’re thinking: AI is well on the way to transforming education for teachers, students, and technology companies!