Xinhua New’s Interview with Steve Deng
Steve Deng: there may be a breakthrough in energizing traditional industries
MATRIX’s Chief AI Scientist, Steve Deng, was recently interviewed by China’s Xinhua News, and shared his insights about AI development. Xinhua News is China’s most influential media agency, run by the Chinese government.
We’ve translated the article into English for your convenience.
Recently, the China Artificial Intelligence Development Report 2018 reported that the AI (artificial intelligence) market size of China in 2017 reached 23.7 billion yuan, with a year-on-year increase of 67%, while the 2018 market is expected to rise by 75%. Steve Deng told a reporter of Xinhua that China has advantages in terms of AI data size, sample size, marked cost and potential talent reserves. Future opportunities of relevant industries will depend primarily on manufacturing.
Steve Deng explained that the most successful applications of AI are image and language processing. Image processing includes various video monitoring, (such as traffic image classification), automatic driving (especially the Tesla driver assistance system), medical image application and face recognition. While language processing includes voice recognition and automatic interpretation.
The next breakthrough may comes from many areas. The first is industrial intelligence, that is, terminal skills combined with sensors and the internet, and the transformation of traditional manufacturing triggered by industrial big data — AI will energize traditional industries. The second is precise cancer treatment, which combines genes, microbial population and other information sources. The third is an application which combines AI and blockchain. The fourth is national defense applications, including new reconnaissance methods based on images and audio signals.
The China Artificial Intelligence Development Report 2018 indicates that China ranks the second in the world in number of AI enterprises, with Beijing as the global hub for AI companies. The number of AI enterprises in China began to explode in 2012, and the number of AI enterprises in China has increased to 1,011 as of June 2018. However, there is a significant difference between China and the U.S. (with a number of 2,028), in this regard.
According to Steve Deng, this surge of AI development comes from three bases: firstly, the improvement of social digitalization allows us to collect data at an unprecedented scale and speed; second, the improvement in computing capabilities, especially the emergence of image processors, has enabled many computationally intensive AI applications to complete training and inference in a time people cannot; third, the development in machine learning and neuroscience over the past six decades has made deep neural networks and other algorithms more advanced.
With regards to the technical threshold to be bridged, Steve Deng says that current AI was supposed to be “narrow AI”, that is, it was meant to solve specific problems. Some recent research indicates that human brains may adopt a predictive processing mechanism, which means that the human brain forms a predictive model according to experience and all a human’s active and passive perception processes are completed under the guidance of this model. Currently, the attention-based model fits this predictive processing mechanism to a certain extent, but it is far from being fully realized.
The human brain is good at dealing with uncertainty, and at present, successful AI technology is always applied to areas with extremely strong certainties (such as AlphaGo aiming at weiqi (go)). Currently, it is widely believed that the most effective approach to deal with uncertainties is with the Bayesian Analysis method, and many scholars believe the human brain achieves is Bayesian reasoning, but its neurological basis is still not entirely clear.
According to the report, AI has been applied to many practical fields including medicine, health, finance, education and security. In 2017, the size of the global robot market reached 23.2 billion U.S. dollars, of which China accounted for 27%.
“There is an excellent environment for artificial intelligence in China in terms of practical need, data samples, talent reserves and investor engagement. The environment for growth in artificial intelligence is also very rich. For example, because of the global need for more advanced digitalized manufacturing, with China’s huge manufacturing industry, the production efficiency between China and European and the US is shrinking. The use of AI technology to achieve a deep digital transformation, combined with the development of smart sensors and terminal processing chips, will bring us tremendous opportunities. ” said Steve Deng.