Parlez-vous artificial intelligence? Two new research papers detail unsupervised machine-learning methods that can do language translation without dictionaries, as reported in Science. The methods also work without parallel text, or identical text that already exists in another language.
The papers, completed independently of one another, use similar methods. Both projects start by building bilingual dictionaries without the aid of a human to say whether they were right or not. Each takes advantage of the fact that relationships between certain words, like tree and leaves or shoes and socks, are similar across languages. This lets the AI look at clusters and connections from one language and learn about how another language works.
When it comes to translating sentences, the new dictionaries are put to the test with some additional help from two methods called back translation and denoising. Back translation converts one sentence to the new language before translating it back. If it doesn’t match the original sentence, the AI tweaks its next attempt and tries to get closer. Denoising works similarly, but moves or takes out a word here or there to keep the AI learning useful structure instead of just copying sentences.
Improving language translation has been a goal for companies like Google and Facebook, with some recent successes. Other attempts, like Google’s recent Pixel ear buds that are meant to translate on the fly, are still a work in progress.
Yann LeCun has a bold new vision for the future of AI
One of the godfathers of deep learning pulls together old ideas to sketch out a fresh path for AI, but raises as many questions as he answers.
Inside a radical new project to democratize AI
A group of over 1,000 AI researchers has created a multilingual large language model bigger than GPT-3—and they’re giving it out for free.
Sony’s racing AI destroyed its human competitors by being nice (and fast)
What Gran Turismo Sophy learned on the racetrack could help shape the future of machines that can work alongside humans, or join us on the roads.
DeepMind has predicted the structure of almost every protein known to science
And it’s giving the data away for free, which could spur new scientific discoveries.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.