Duolingo, the popular free language-learning platform, launched three language “tutors” in its iOS app today, marketing the AI-powered chatbots as a way of helping users learn a new language without having to pay for expensive lessons. Users can text back and forth to practice a new language, giving learners an eternally patient, nonjudgmental, on-demand instructor. Chatbots, the services you interact with via a chat interface like Facebook messenger, have been hailed as the next big thing in tech. People are now using messenger apps more than social media networks, which means app developers are increasingly turning to bots to help their users access a wide range of services. Major companies like Facebook and Microsoft are trying to capitalize on that potential by rolling out platforms that make it easer for developers to build chatbots, and Google made a virtual assistant accessible through chatbot the central feature of its new phone. Despite the hype, chatbots have fallen disappointingly short. Last March, Twitter got Microsoft’s artificial intelligence chatbot “Tay” to come to the dark side in under 24 hours. Tay started the day as a peaceful, human-loving bot and ended up spewing a barrage of racist tweets in a disturbingly short amount of time. Facebook hasn’t had much luck with its bots either, despite Mark Zuckerberg’s mission to build an “AI to answer any question you have.” The Messenger chatbots have been criticized for being terrible conversationalists, prone to long stretches of silence and regurgitating predetermined answers. Duolingo’s chatbots are Renèe the Driver, Chef Roberto, and Officer Ada, with the promise of more characters coming to the app soon. Learners can use them to practice French, Spanish, and German, and more languages will be added according to demand. The bots add a personal touch to Duolingo’s language programs, and they’re powered by AI, so the more you interact with them the smarter they get. Translation is also a very well-defined domain of expertise, meaning these bots are likely to make fewer mistakes than AI aimed at open-ended conversation.
