If AI is so good, why are there still so many jobs for translators?

Episode Summary

Executive Summary: The episode examines AI’s impact on jobs through the lens of translation, a field where machine performance is already strong and steadily improving. Duolingo CEO Luis von Ahn argues AI is augmenting rather than instantly replacing human translators, creating a hybrid workflow that boosts productivity while preserving humans for high-stakes, nuanced tasks. He’s optimistic about AI’s benefits but worries about a too-fast transition.

Main Topics: AI and the translation job market: The discussion centers on whether AI is eliminating translator/interpreter jobs or reshaping them. Von Ahn says computer translation has improved for years, so the change is gradual rather than abrupt. Human-AI hybrid workflows: Duolingo uses AI for bulk translation work and humans for review, consistency, and sensitive UI text, illustrating a mixed model rather than full automation. Duolingo’s AI adoption and product strategy: Duolingo has used AI from the start, but large language models unlocked new capabilities like conversation practice and faster content creation. The contractor layoff controversy: Von Ahn rejects media claims that Duolingo carried out major layoffs, saying non-renewal of some contractors reflected changing workload and increased automation. Translation as a proxy for AI progress: Greg uses translation to gauge whether AI hype matches reality, since it is one of the areas where AI is already relatively advanced. Optimism and caution about AI’s social impact: Von Ahn is broadly optimistic about AI’s long-term potential, but anxious about job displacement if the transition happens too quickly.

Key Arguments: Computer translation is not a sudden breakthrough; it has been improving for over a decade, so job change is gradual rather than a single wipeout event. AI increases demand for translation because cheaper and faster translation expands use cases, which can offset some labor displacement. Humans are still needed for high-stakes, nuanced, or brand-sensitive translation, especially in legal, military, and user-interface contexts. Duolingo’s content translation is mostly automated, but UI copy like buttons requires humans to ensure consistency and avoid costly mistakes. The contractor reductions at Duolingo were not full layoffs of employees; they reflected reduced need for some contractor work as automation improved. Large language models enabled Duolingo to add conversation practice, a capability it previously could not support. AI is likely to raise productivity dramatically, allowing a small team to do work that once required many people over years. A slower transition to AI is preferable because rapid job losses could create serious social instability.

Data Points: Projected growth in U.S. translators/interpreters: 4% over the next decade - Cited from the U.S. Bureau of Labor Statistics as roughly equal to average occupational growth Duolingo contractor force: a few hundred contractors - Von Ahn describes the size of the contractor pool before some contracts were not renewed Remaining translators at Duolingo: dozens - Von Ahn says Duolingo still employs a small number of translators AI content creation time reduction: 5 years to about 3 months - A proposed animated language-learning feature that once seemed too expensive became feasible with AI Productivity gain: 10x - Von Ahn describes the overall productivity uplift from AI-enabled workflows GPT-4 access timing: almost two years ago / a little less than two years ago - Duolingo began experimenting with GPT-4 before official feature launches AI hype reference: 60 years or longer - Von Ahn notes that the concept of AI predates the current LLM wave by decades Workforce impact claim: 0 full-time employees affected - Von Ahn says the contractor changes did not involve layoffs of full-time staff

Pivotal Quotes: "This is a gradual process which started out years and years ago." — Luis von Ahn: Explaining why AI translation won’t suddenly eliminate translator jobs overnight "We just want to make sure that it's all consistent, that it has the same voice, the playful voice, etc. And so, for that, we still employ humans." — Luis von Ahn: Describing why Duolingo still uses human translators for user-facing text "I'm mostly optimistic. I'm mostly optimistic. But I am anxious." — Luis von Ahn: Summarizing his overall view of AI’s promise and risks

Implications: AI is likely to reshape jobs by expanding output and lowering costs before fully replacing workers. For listeners, the key lesson is that the most immediate effect may be hybrid human-AI work, not instant automation.