AI research opens new opportunities

After announcing his retirement from Microsoft in November, Harry Shum, the former vice-president of artificial intelligence and research at Microsoft, is back in the spotlight.

At an online appointment ceremony on Thursday, Shum was appointed as adjunct professor at Tsinghua University. It is the first time that Tsinghua held such a virtual appointment due to the public health crisis, which has claimed more than 3,000 lives and infected over 80,000 people across the country.

"The sudden outbreak of the novel coronavirus is a challenge faced by all mankind.… We should gather talents from all over the world to solve our common problems with an open mind, a rigorous attitude toward learning and a spirit of perseverance," Shum said at the ceremony.

Outside of China, the virus was reported in 101 countries and regions as of Sunday, according to WHO.

Qiu Yong, president of Tsinghua, said that artificial intelligence will bring tremendous changes to human society, and its influence on each of us will go beyond our imagination.

"We believe that having world-class experts like Shum joining us will not only have a positive influence on Tsinghua University but also contribute to the development of artificial intelligence in China and the world," said Qiu.

He presented the appointment letter at Tsinghua University to Shum who was at the Global Innovation eXchange Institute in Seattle. In 2015, the GIX was established by Tsinghua University and the University of Washington with the initial foundational support from Microsoft.

While artificial intelligence has been part of our life, from online shopping recommendations, stock picks, digital assistant like Siri to medical diagnoses, Shum warned that "AI is already making decisions we don't understand".

His remarks came at a livestreaming lecture titled Engineering Responsible AI after the appointment ceremony.

Shum, 53, said that more and more AI is like a "blackbox", which researchers need to "dive in" to explain how algorithms work. Known as black-box models, they are defined as systems in which the journey from input to output is next to impossible for even their developers to comprehend.

It is an emerging field called explainable AI to help make users and customers understand what is going on and trust the technology.

"Explainable AI will be very important for the next decade," Shum said.

With explainable AI, engineers can use it in testing to find out biased decisions and discover an emerging bias, which is a serious challenge posed by deep learning models.

"It might have inherited some bias because of the data we use to train it. Such biases, if detected, can be corrected with additional training", Shum said.

A MIT study in 2017 called Gender Shades showed bias in some of the most used facial recognition systems from Microsoft and IBM. They were better at identifying the gender of white men's faces than with the gender of darker-skinned or female faces.

A Microsoft study also showed that its system trained on web data would form connections between "he" and "programmer" or "she" and "homemaker".

Shum said the machine learning system is not as accurate as engineers would like to see.

"In fact we are the first generation of human being ever living with AI. … Like or not, we have no choices. We will live with AI and AI will get stronger and stronger. But we can decide how we are going to build AI and how we are going to use AI," Shum concluded.

According to Forbes, about 75 countries are using AI technology for city management through smart city platforms, facial recognition systems and smart policing.

In investment terms, the percentage of companies spending over $50 million on big data and AI increased to 64.8 percent this year from just 39.7 percent in 2018.

He left an open question at the end: "Can we accept a future in which AI is making decisions we cannot explain?"

With no Chinese subtitles, the lecture in English has attracted more than 700,000 viewers.

A netizen called SIMOS commented that, "In the current 'artificial intelligence', it is more 'artificial' than 'intelligence'. We still have a long way to go."

Shum spent 23 years at Microsoft. Born in Nanjing, Jiangsu province, in 1966 and with a doctorate in robotics from Carnegie Mellon University, he is one of the few Chinese computer scientists widely known in the tech world in the United States.