AlphaGo’s landslide victory over Korean Go grandmaster Lee Se-Dol has stirred up concerns that AI might spell the end of the human race. Several leading figures like Stephen Hawking, Bill Gates and Elon Musk, echoed the voice, further raising people’s worries for the existential threat.
“The panic was brought up again after the Go competition because evidence show that the results of deep learning are a bit uncontrollable for human perceptions.” said Xu Bing, cofounder and VP at SenseTime. “AlphaGo has made several incomprehensible moves in the game against Lee. Human Go masters consider them as “bad” moves, but they turned out to be decisive steps that lead to its victory.”
However, the three AI experts spoke at TechCrunch Beijing this Monday agreed that public worries for this problem is over exaggerated because technologically speaking there’s still tons of obstacles for AI products to overcome.
“AI technologies is based on deep-learning of human brains, which has more than 300 billion neurons. Currently, no computer platform can accommodate this number of parameters.” said Xu Bing.
Furthermore, AI technologies must own people’s creativity powers if they want to outperform human race, according to Du Yujin, director of R&D at Microsoft Cortana. “Robots could do great jobs in face or voice recognition, but only in fields that have already been defined by human minds.”
In the future five to ten years, the focus of the industry is how to improve efficiency and increase productivity of human labor, like driving, face and voice recognition. In this period, AI may outperform human in certain fields, but certainly not at a comprehensive level, said Yang Ming, co-founder and VP at Horizon Robotics.
”In addition, human’s knowledge is accumulated through a long learning curve, but AI technologies has yet to conquer the long-term memories.” Yang said.