President Donald Trump ca n’t speak Chinese, he didn’t even go so far as to say that artificial intelligence is changing the world and that iFlyTek is great. Yet on the sidelines of a technology conference in 2017, Trump appeared on a big screen, welcomed attendees in English, and finally spoke fluently in Mandarin Chinese.
This little “miracle” was the fruit of the work of the Chinese company iFlyTek , which had conceived a digital model of Trump’s voice so perfect that it mimicked not only his tone and intonation, but also his speech pattern. And all this thanks to the extraordinary training of its algorithms.
While Trump’s speech was not totally in sync with the movement of his lips, the fact that congressional attendees could hear the American president addressing them in perfect Mandarin overshadowed the small detail. Without a doubt, it was a technological milestone : allowing anyone to say anything in any language while preserving their voice and the way they speak.
Not surprisingly, this Chinese company has already received distinctions in various international artificial intelligence competitions where the benefits of voice recognition and synthesis, image recognition and automatic translation were put to the test.
Other presidential speeches
It is not the only example in which iFlyTek demonstrated its competence. Also former president Barack Obama has spoken Chinese thanks to the same technique. What he said roughly was: “I have learned Chinese with the help of iFlytek. I think my Chinese is better than Trump’s. What do you think? “
The opinion should be favorable, because iFlyTek has repeatedly outperformed giants like Google, Facebook or IBM in the field of natural language processing (the ability to decipher general meaning rather than just a few words). This is possible because you have to deal with a particularly complex language in the field of speech recognition and synthesis: Chinese is a tonal language, that is, the same word can mean different things depending on the tone with which it is pronounced.
This requires iFlyTek to need to capture more information and separate it into two separate streams to improve reliability, which indirectly supports the development of speech recognition systems under noise and other distortion conditions.
The company was founded by Liu Qingfeng and five other students from China University of Science and Technology in 1999. Since then, iFlyTek has become the world’s most valuable voice technology company. In 2018, the company signed a five-year collaboration agreement with MIT’s Computer Science and Artificial Intelligence Laboratory. Last year, it launched systems that allow people to control their cars, televisions and appliances through voice; and has also established a joint venture to develop educational products incorporating its simultaneous translation features.
At the moment, iFlyTek owns 80 percent of the market share of speech recognition technology in the People’s Republic of China. At this rate, it won’t take long for us all to say “ni hao”.
