Researchers from Microsoft recently announced they’ve created the first deep learning translation system capable of translating sentences of news articles from Chinese to English with the same level of accuracy as a person.
Microsoft used NVIDIA Tesla GPUs and millions of sentences from various online newspapers to train their neural network. The team used a dual learning system where the AI would reverse the translation to see how well it did.
Once trained, the researchers used GPUs on the Microsoft Azure cloud for inference which was able to achieve human parity.
To verify the accuracy of the results, Microsoft hired independent bilingual language consultants.
“Hitting human parity in a machine translation task is a dream that all of us have had,”Xuedong Huang, a technical Microsoft fellow said in a company blog. “We just didn’t realize we’d be able to hit it so soon,” he added.
One of the reasons translation is such a challenge is because there’s rarely one right way to do it.
The team cautioned that the breakthrough does not mean that machine translation is a solved problem. However, the techniques learned in the test will be useful for improving machine translation in other languages, and that they could be used to make other AI breakthroughs, the researchers said.
Read more >
AI-Generated Summary
- Researchers from Microsoft created a deep learning translation system that translates Chinese news articles to English with human-like accuracy using NVIDIA Tesla GPUs and millions of sentences from online newspapers.
- The system was trained using a dual learning method where the AI reversed the translation to test its accuracy, and was then verified by independent bilingual language consultants.
- The breakthrough is expected to improve machine translation in other languages and potentially lead to other AI advancements, although the researchers caution that machine translation is not yet a solved problem.
AI-generated content may summarize information incompletely. Verify important information. Learn more
About the Authors
