In August 2017, Microsoft reported that its speech recognition AI can transcribe as well as humans. The statistics said that the AI has achieved an error rate of 5.1% – a level that matches the error rate of expert human transcribers. This 2018, the company has reached another breakthrough.
This 2018, Microsoft makes another breakthrough in Natural Language Processing (NLP) as their machine-translation system can now translate Chinese news articles to English with the same accuracy of human translators. This is important news, not only because 16% of the world’s population speaks Chinese, or because China is making waves in AI development. But also due to the fact that Chinese is one of the most difficult languages for machines to learn. Moreover, data in Chinese is overwhelming, thanks to the 70+ million Chinese people who use the internet every day.
The public can now test the Microsoft’s online translator but the company is not yet ready to make this development commercially available.
In the article below, Microsoft explains how it was able to train its AI using deep neural networks, and other training techniques. Through this development, it becomes easier to see that the technology available is making it so much easier for human-level AI to become a reality.
[contentcards url=”https://blogs.microsoft.com/ai/machine-translation-news-test-set-human-parity/” target=”_blank”]
You may also like
-
WEF Launches Metaverse, Predicts Digital Lives Will Become More Meaningful Than Physical Lives
-
China Deploys Facial Recognition To Limit Playing Games Among Youth
-
Digital: Is this the kind of future you want?
-
Human augmentation: Making your brain hackable
-
Augmented reality pioneers warn against Metaverse, says it could be worse than social media