Forget about “Scandal in the simultaneous interpretation” – the biggest crisis for Iflytek lies not only in technology, but also in its business and its ethics. So let’s not stick to technical side only, but give a talk about the business model instead.
On September 20th, a simultaneous interpreter in Shanghai blogged a message, reporting that at the 2018 Innovation and Emerging Industries Development (IEID) held on that same day, the Iflytek system deployed on site did not interpret the speech at all, but merely used speech recognition to transcribe the simultaneous interpretation by human translators on the screen.
All of a sudden a public outcry was provoked. Some said that Iflytek was “cheating in simultaneous interpretation”, saying it was interpreted by AI which was not – it was basically false propaganda; some explained that it was “human-machine doubling”, i.e. the interpretation is a combination of human and machine; some also said that machine translation technology was still in its infancy, AI simultaneous interpretation is unreliable. When you need a language translation for serious purpose, you’d still better to get a reliable human translation service provider.
We can say that whether the company explains that it is human-machine coupling or voice recognition, Iflytek’s being caught up in this vortex of public opinion is due to the fact that it deliberately induced the audience in the early stage of publicity, making the public think that Iflytek’s interpretation is the result of AI. Actually it is someone digs the pit, but finally buries oneself.
As we said, the biggest crisis for Iflytek lies not only in technology, but also in its current business model.
After reading the introduction of the interpretation products of Iflytek on various occasions and the promotion of the products in different BBS and conferences, it is easy to get the impression that the AI interpretation of Iflytek can replace simultaneous interpretation. To tell the truth, many people in the interpretation industry think so too, and are convinced that they will lose their jobs in the near future, and this has even affected the recruitment of many foreign language majors.
However, the reality is harsh. Technological innovation is by no means to be opportunistic; it is never to promote something to 100 which is in fact 50. The more so, the easier it is to bite back. For Iflytek, If you can’t achieve automatic machine interpretation without human help, if you must use human-machine coupling, then make it clear, instead of vaguely telling the audience in the media that you have reached this level of technology.
Turning speech into words (speech recognition) and automatic machine interpretation (AI) are completely different on the degree of difficulty. There is no room for sloppiness in technical matters. As a company born in the university, Iflytek should take a serious approach to academics and technology. Especially in publicity, you can’t just say one when actually it’s ten, and let everyone think that your one is ten; thus if the remaining nine is not pointed out by others, you can enjoy the fruits of ten.
Iflytek’s statement missed the point.
In the same evening, Iflytek issued an explaining statement for the above “cheating in simultaneous interpretation” incident. I have read it, and it can be conclude to two points: first, the technology adopted in the conference held on September 20th was human-machine coupling; second, at the World Artificial Intelligence Conference held on September 17th before that, the interpreting mode of Iflytek showed great performance.
These two points are quite humorous.
If the interpreters had already been told it was the combination of machine and human, I believe there won’t be someone so angry to make a statement, saying that Iflytek plagiarized the result of simultaneous interpreters. The truth is, the interpreters themselves have no idea of human-machine coupling at all. Iflytek should take complete responsibility. Either the work is not done properly, or it is deliberately creating a fuzzy space, making people who don’t know the truth assume that the interpretation on site is all the work of Iflytek’s AI robot.
For the second point, as long as you have attended the conference on 17th, or watched the video of the conference, you will fully feel the level of Iflytek’s interpretation – we can say it’s quite poor. Also in the Sino-Russian Industrial Competition a few days ago, some of the guests were dumbfounded by the bizarre scene when Iflytek heard the Russian simultaneous interpretation on site. In the present situation, it is quite difficult to interpret some professional speeches with the technology of Iflytek. There is no shame in admitting this generously.
It’s better to do something real about it instead of exchanging fire verbally.
Compared with Iflytek’s performance in the capital market, this “cheating in simultaneous interpretation” is really nothing. As of Friday, the share stock price of Iflytek (002230) had fallen 40 percent from its peak in 2017 and was now worth ￥63 billion. But in terms of valuation, with PE ratio of 138 times and PB of 8.1 times, it’s still very expensive!
Instead of quarreling with netizens over AI or human-machine coupling, or paying more attention to media, the company should honestly think about how to make their business models more reliable.
From 2013 to 2017, Iflytek has issued additional financing every year except 2014. It has raised more than ￥4 billion from the market. However, it has paid out just over ￥600 million for cumulative dividends over these years. As I write this article, Iflytek has again passed the additional issuance plan for 2018, and intends to raise several billion – larger and larger amount.
A-share investors cannot always pay for a company’s dreams; if the technology is really so good, why not do something practical and give investors some feedbacks?
For an investor, whether it is human-machine coupling or AI, it is all surface work; all in all, it only matters whether the business model of Iflytek can succeed. But for now, the lack of core competitiveness, the lack of ability to commercialize and the lack of ecological construction capacity are the inherent problems of Iflytek. At present, there is still no signs of improvement for these problems.
China’s AI industry still has a long way to go
Kai-fu Lee said that China’s AI industry would surpass that of the United States in five years. I don’t think it’s important to exceed the US or not. The key is to get things done for now. Iflytek has always been a good company in my mind. However, if there is any deviation in the reports, the company should take its responsibilities instead of pushing it all to the reporters.
As we said before, in recent years, the concept of AI was super popular, therefore many companies took this theme to the capital market to tell stories and blow the valuation. Companies with a little bit of technology were valued at more than ￥100 million before they had even been established. But is it really good for the industry? Coming forth with a rush can only result in chaos in the end. I really hope that China’s AI industry can develop in a down-to-earth way for some years and do a good job in basic research, instead of just promoting concepts.
We understand that Iflytek did do a lot of work, but you absolutely cannot take some functions that you haven’t accomplished yet and make people mistakenly think that you’ve already accomplished it through some kind of fuzzy operation. Why there are so many bubbles in China’s AI industry? It’s because a lot of companies and entrepreneurs take the functions not yet being achieved to call their own, as well as put other people’s things into a shell to call their own. As a famous company in the field of speech recognition, Iflytek should not make such mistakes and should set higher standards for itself.