OPINION: AI realities discussed at Economist event in Hong Kong

Anyone in the events business should attend a full-day event by The Economist. The respected UK publication—published continuously since 1843—does it right: editors serve as moderators and guests discuss tech in real-world terms.

PowerPoint presentations are forbidden at Economist events. Speakers do not display complex tech diagrams while audience members zone out on their smartphones. Instead, they speak and people listen.

The Economist hosted an all-day event last week in Hong Kong: “Building the intelligent company.” Here's an overview.

Discussion is valuable
The aim: “valuable discussion on how (and why) companies harness the potential of AI”—promoting the value of discussion as opposed to the PowerPoint barrage. How many times have you heard a speaker say: “well I don't have time for the rest of the slides”? 

Our hyperconnected world wants us to pay more attention to our data-spewing gadgets than listening to people discuss an issue. But when you assemble panel with people like Frank Tong (global head of innovation and strategic investments, HSBC), Ross McCullough (president, Asia-Pacific, UPS) and Steve Monaghan (chief executive, GenLife), the realities of banking, delivery logistics, and insurance provide essential perspective on this phenomenon called AI.

“We do 20 million deliveries daily so we rely on AI to speed and scale,” said McCullough. “Humans can't do it.” The UPS exec also said that enhanced data-collation gives UPS “flexibility on both delivery times and [the importance of] package contents.”

“Underwriting [insurance] is math-heavy, so AI is useful for that,” said Monaghan. The insurance executive said that if his firm could use data pertaining to early-detection for cancer and other proactive measures, then “we could help mitigate risk.”

Tong from HSBC said that big data and AI “poses the majority of problems I deal with at the moment.” He also cited voice analytics, prompting an intriguing question from an audience member: why do we still have a voice-menu saying “Press 1 for Cantonese, 2 for English” if voice analytics can handle that task?

Tong's answer was equally intriguing: he said that while the technology does well with English and Mandarin, Cantonese continues to confuse the listening bots. Clearly, AI is not an exact science—in fact, when one of the editor-moderators asked the audience if they agreed with the phrase “AI is advanced automation,” only one of the audience of 180 dissented.

“General purpose AI hasn't hit the mainstream yet and is 10-15 years away,” said Monaghan. “AI is 'less-wrong'— it's not right all the time.”

He said that adoption within his industry is imperative. “Most healthcare systems will go bankrupt within the next 10-15 years because of aging populations,” he said, “unless we use the 'force-multiplier' of AI scanning X-rays.” Such a system would not replace human doctors, but would help prioritize workloads.

The ethics of AI
The ethical considerations of AI usage permeated discussions throughout the day. Economist editor Tom Standage (both he and fellow editor Vijay Vaitheeswaran served as exemplary moderators) cited a Rand study on driverless-car deployment. For example, said Standage (I paraphrase here): if the autonomous vehicles are ten times safer, thus reducing traffic accidents tenfold, will it be accepted by the general public considering that mainstream media will trumpet every “driverless car accident” to help drive their page-views? If not tenfold, how about a hundredfold?

These questions remain unanswered, but a discussion titled “Regulating AI: Better safe than sorry?” brought yet more experts to the stage, including Andy Chun, associate professor at the City University of Hong Kong. Joining Chun were Jennifer Van Dale (a partner at law firm Eversheds Sunderland), Tze Yun Leong (director of AI technology, AI Singapore), and Naveen Menon, president for Southeast Asia at Cisco.

Van Dale said that with facial recognition technology, “ethnicity is an issue, so we need standards. Does that fall under existing non-discrimination laws, or do we need new laws?”

Menon said that some degree of regulation is desirable to help create a “level playing-field” and competition without monopolies. He emphasized the need for data protection.

“I'm against regulating AI, but certifying cars is a good idea,” said Chun. “We as humans assume that AI is 'smart' and moving forward, but so many companies will be releasing AI products that we don't really know.”

Van Dale said that in the case of autonomous cars, questions like 'why did the car go left rather than right?' will arise. “Sometimes,” she said, “that data is proprietary.” Such situations bring a host of associated legal issues.

The Great Debate
One of the more entertaining segments of the full-day program was a debate on AI job-creation versus destruction. Arguing for job creation was Xania Wong, founder and chief executive of JOBDOH, a former Cyberport incubatee which now uses an “algorithmic driven hiring platform [that] connects employers with curated on-demand talents in critical time frame.” Arguing for job destruction was Gerardo Salandra, chief executive and founder of chatbot provider Rocketbots and chairman of the Artificial Intelligence Society of Hong Kong. In the middle was stalwart Economist editor Vijay Vaitheeswaran who served as moderator and timekeeper.

Both debaters are businesspeople and not professional arguers, which made the debate less structured and more interesting. Questions from the floor varied in relevance, but Vaitheeswaran ran a tight ship and attempted to achieve consensus.

My consensus? Events provide better value when PowerPoint presentations are banned. As for AI, the consensus was produced by the audience: it's best described as “advanced automation”...at least in 2018.