A girl touching the robot Sophia, dressed in the Korean traditional costume, during the Fourth Industrial Revolution and AI Robots exhibition in Seoul. Photo: Yonhap
Mind the Gap
by Peter Guy
Mind the Gap
by Peter Guy

The biggest limitation of artificial intelligence is it’s only as smart as the data sets served

AI’s main limitation is that it learns from given data. There is no other way that knowledge can be integrated, unlike human learning. This means that any inaccuracies in the data will be reflected in the results.

“The mission is too important for me to allow you to jeopardise it,” said Hal the computer to the astronauts in 2001: A Space Odyssey.

The artificial intelligence (AI) race has taken on the portentous public relations spectre of the Soviet-USA race to the moon, or to build the first atomic bomb. Periodic announcements of milestone achievements confuse many because few understand what it means for an algorithm to pass a test reading literature.

The dominance of big American tech companies- Facebook, Amazon, Netflix, Uber and Google versus the rise of China’s biggest tech outfits: Baidu, Alibaba, Tencent, has created intense competition for resources and specialists. Yet, what are all of them racing towards?

Ignore the science fiction scenarios of dystopia; like the worry of AI causing the next war, or financial crash that will impoverish or extinguish humans.

Fear of the unknown fuels the imagination and dark humour. Just before the test of the first atomic bomb at Trinity, nuclear scientist Enrico Fermi offered side bets to colleagues on whether the explosion would incinerate the earth’s atmosphere.

The reality is that the impact of AI will be most profound in specific business applications. The main limitation is that it learns from given data. There is no other way that knowledge can be integrated, unlike human learning. This means that any inaccuracies in the data will be reflected in the results. Today’s AI systems are trained to do a clearly defined task. The system that plays chess cannot play solitaire or poker.

In health care, AI applications can provide personalised medicine and diagnostic readings. A personal health care assistant or “bot” can act as a lifestyle adviser, reminding you to take your medication. AI can support virtual shopping capabilities perhaps a “Shazam” for retailing.

Shazam is an app that can identify music, movies, ads and movies based on being given a short sample played to the device. Point your smartphone’s camera at something you wish to buy, and as long as it can be bought, like a friend’s jacket, AI can scour the world and help you buy it at the best price.

The greater application of automation used to only affect the fringes of the economy’s work force - reducing the need for manual labour, substituting it with capital and technology. But now, AI threatens the core operations of many organisations - analysis and decision making that was once the competency of managers.

By offering automation, intelligence, learning and accuracy beyond human ability, AI enhances the value of data beyond what humans can accomplish. So fewer humans are needed.

Technology recruiters in San Francisco tell me that AI specialists - newly minted Ph.D graduates and candidates with fewer qualifications and two or three years of experience, can make between US$300,000 to US$500,000 a year, or more, in salary and share options.

Highly experienced AI specialists who have a proven track record in managing projects earn more than top bankers. In a 2017 court filing over an intellectual property suit with Uber, Google revealed that one of the leaders of its self-driving car department, Anthony Levandowski, received US$120 million in pay and incentives since 2007.

Fewer than 10,000 people have the skill sets required to effectively work on serious artificial intelligence research, according to estimate by Element Ai, an independent lab in Montreal. Human resource constraints may be the ultimate limitation for successful AI development.

But I still look forward to asking the first Chinese AI systems: “What happened on June 4, 1989 in Tiananmen Square?” or “Name the top 100 corrupt officials in China.”

After all, a true AI system objectively studies all available data and cannot lie.

Peter Guy is a financial writer and former international banker.

This article appeared in the South China Morning Post print edition as: AI is only as smart as the given data input