简体中文
Try typing“the machines”into Google and chances are that one of the top results the artificial intelligence-powered search engine will return is the phrase: “The Machines are Coming”.
After a 2016 filled with high-profile advances in artificial intelligence (AI), leading technologists say this could be a breakout year in the development of intelligent machines that emulate humans.
Asia, until now lagging Silicon Valley in AI, will play a bigger role as the field cements itself at the pinnacle of the technology world in 2017, the experts say.
AI – technically, a computing field that involves the analysis of large troves of data to predict outcomes and patterns – is as old as modern computers but its esoteric nature means it has long endured caricatures of its actual potential – think for example, the 1960s space age cartoonThe Jetsons, which featured a sentient robot maid and automated flying cars (both of which we are still waiting for, even 50 years on).
A model of Rosie the Robot from The Jetsons. Fifty years on, we’re still waiting for sentient robot maids. Photo: AFP
Now, a confluence of factors has given rise to hopes that computers with human-like cognitive ability may soon be a reality.
That sentiment was etched in public consciousness last March when AlphaGo, a sophisticated Google AI platform, staged a stunning victory against a human grandmaster in the ancient Chinese game of Go – seen until then as beyond the comprehension of even the most advanced computers.
And with billions of dollars being sunk into AI research by Silicon Valley’s biggest players as well as China’s big three “BAT” technology firms – Baidu, Alibaba and Tencent – coupled with an explosion in data availability and vital computing power, the field has permanently been lifted out of the computing wilderness.
AI researchers say the technology has the potential to revolutionise almost every aspect of the global economy and human life – from the detection and treatment of cancer to the way management consulting and investment banking is conducted.
“What we can expect in the next few years is that the technology that drives self-driving cars and the same technology that drives AlphaGo will be consolidated in different fields such as transportation and health care,” said Soumith Chintala, a research engineer with Facebook AI Research.
In 2016, delirium over AI was mainly due to advances in the subfields of “machine learning” and “deep learning”.
South Korean Go grandmaster Lee Sedol loses his final match to AlphaGo, sealing the AI system’s 4:1 victory. Photo: AP
Machine learning is the process where an algorithm is trained to perform tasks that it is not explicitly programmed for, utilising a large amount of data related to that task. Deep learning technology is a subset of machine learning. It aims to mimic neural networks in the human brain, passing large amounts of data through matrices of artificial neurons where it is processed and analysed at hyper-fast speeds.
Pascale Fung, an AI researcher at the Hong Kong University of Science and Technology (HKUST), said several milestones have been reached in developing computers that are similar to the human brain. Speech and emotional recognition were among the areas “reaching new milestones”, Fung said.
Asia-focused AI experts say the region has lagged the West in research, but its technology companies and universities have enormous potential to make up for lost ground.
Wu Haishan, a senior Baidu data scientist, at the Baidu Technology Park in Beijing, China. Baidu is widely seen as being at the forefront of AI in Asia. Photo: Bloomberg
Baidu, China’s top search engine, is widely seen as being at the forefront of AI in Asia. It has AI labs in Silicon Valley and Beijing. In 2014 it poached Andrew Ng, the son of Hong Kong parents who previously led Google’s AI operations and co-founded the online learning platform Coursera.
“The BAT companies know they will have to compete with the likes of Google and Facebook for talent in AI, and we will likely see this heat up in the coming years,” said Tak Lo, the Hong Kong-based managing director of Zeroth.ai, an accelerator programme for Asian AI start-ups.
Start-up ecosystems that existed in national silos in countries like China, Japan and South Korea as opposed to the contiguous landscape in the West were a stumbling block for AI growth in Asia, Tak said.
Facebook’s Chintala, who is of Indian descent, said Asians made up a significant part of the global AI and deep learning research community. “As individuals, we are already making good impact in the field,” he said.
But research in places like India tended “to be of a lower bar than research in top-tier American or European universities or labs,” said the New York-based researcher.
“This is slowly but steadily changing. The key to closing this gap is to cultivate good mentorship programmes for young researchers, and that’s happening rapidly.”
Leading Asian hubs Singapore and Hong Kong have seen the hatching of several deep learning-focused start-ups amid a surge in demand from regional businesses for services powered by the technology.
Technology data and research firm CB Insights said in December that 137 AI-linked start-ups had been acquired around the world since 2011, with 40 this year alone.
Among them were firms specialising in AI “chatbots”. These are software based on deep learning technology that enables users to perform tasks such as ordering food, paying a bill or hailing an Uber ride using a text or voice message.
Chatbots are software based on deep learning technology that enables users to perform tasks such as ordering food, paying a bill or hailing an Uber ride (above) using a text or voice message. Photo: Edward Wong
One such firm is Singapore-based Active.ai, which has developed a chatbot that enables bank customers to conduct business by chatting with an AI-powered platform accessed through messaging apps.
“We previously had internet banking, and then an era of mobile banking and ‘mobile first’. Now we are firmly moving into an era of AI banking and ‘AI first’,” said Shankar Narayanan, the firm’s co-founder.
And in Hong Kong, early stage start-up Clare.ai is developing similar intelligent chatbot technology, with Cantonese voice recognition alongside English and other major languages.
“In 2017 we will see further use of such technology by businesses like banks, insurance companies and retail firms. But it will take a few more years before it becomes mainstream,” said Ken Yeung, co-founder of the Hong Kong firm.
Television displays show the defeat of South Korean Go grandmaster Lee Sedol to Google’s artificial intelligence program, AlphaGo at an electronics store in Seoul, South Korea. Photo: AP
Fung, the HKUST researcher, said China, the world’s most vital factory floor, was in a unique position to capitalise on leaps in AI because “collaboration between research, development and manufacturing can be done very fast”.
The treasure trove of data available to China’s BAT internet trinity could also see the companies use “the latest machine learning algorithms to make use of this data to improve their services”.
Amid the dizzying potential in AI, experts are at pains to emphasise the need to contain expectations and dispel fears that the technology could adversely affect human life.
Major US technology companies in September announced the formation of the Partnership on Artificial Intelligence to Benefit People and Society, a non-profit alliance that will conduct research on ethical AI and “advance public understanding and awareness” of the technology.
“It’s simply important to understand what AI is in a fundamental way, because without understanding what it is folks might overestimate the ability of the current AI technology, leading to bad investments and decisions,” said Chintala, the Facebook researcher.
He added that the “unattractive but honest thing to say is that we are [only] on the cusp of incremental AI at the moment”.
Actor Arnold Schwarzenegger on the set of Terminator. Don’t worry, experts brush off fears that AI technology could spawn the arrival of insidious computer networks capable of dominating humans. Image: Corbis
And Nello Cristianini, an AI professor at Britain’s Bristol University, said there was a need for users and regulators to recognise potential abuses of AI technology.
The security of personal data, addiction to adaptive technology, as well as unauthorised leaks of predictions about individuals’ financial behaviour to the likes of insurers and banks are among the key risks linked to AI.
“It would be a good thing if 2017 brought cultural progress in these areas, along with the expected technical one, to avoid a bad awakening when it is too late,” Cristianini said.
The experts are quick to brush off fears that AI technology could spawn the arrival of “Skynet”, the insidious computer network that dominates humans in theTerminatorfilms (above).
“Some of the fears, such as robots taking over humans, are really far-fetched ... we have made a lot of progress and breakthroughs but we are no way near building machines with general intelligence yet,” said Fung.
(SCMP)
