With AI, the human aspect must still come first

AI is one of the next big technological trends that is finding its way into all manner of applications around the world. But as entire countries jump on board, with coding classes being made mandatory at primary school level and older workers urged to learn tech skills to stay current, industry experts are urging people and companies not to neglect the critical human aspect of this technology.

Before good AI comes good ethics

Responsible AI has to become a mainstream thing, said panellists at a General Assembly seminar in Singapore on Thursday.

“It is not possible to separate the ethical responsibility aspect from AI,” said Sunita Kannan, ASEAN Lead for AI Advisory/Strategy and Responsible AI. She pointed out that before a company decides to bring in AI, its own morals must be transparent enough for people to understand the implications of introducing AI; the organisation as a whole must recognize the need to have an ethical wall in place around whatever it is doing with the technology.

“The problems you see with AI today--bias, preferences, selecting certain types of people over others, whether in favor of majorities or minorities--those are all fundamentally human issues,” observed David Robinson, the former Chief Technology Officer of MyRepublic. “We need to put the human element together with the technology to resolve these.”

Research and education are already tackling the issue. The MIT-IBM Watson AI Lab is studying how the human mind selects and applies decision-making principles, and looking into how machines can be made to do the same. Harvard researchers have developed a framework, dubbed “Embedded EthiCS”, for incorporating ethical reasoning into computer science education in such a way that it may become part of the way future products are developed.

And after good ethics comes sound psychology

“We need to start looking at it from a psychological point of view,” said Robinson. “You need that because you are dealing with customers, and to be able to understand human behavior and also to understand data, that’s a very powerful combination.”

He added, however, that this would likely introduce new problems simply because engineers and psychologists are trained to have diverse worldviews and approaches, and getting these two groups to work together as a team would be challenging.

It is not just pure psychology that is needed, however. 

“One inherent skill that cannot be replaced is management,” said Kannan. “By management, I mean the constant review and optimisation of processes and people.” She added that there is a great need for sociology and law skills in the field: sociology because of how important the human element is to develop AI, and law because of the regulatory considerations surrounding data.

“The more diversity, the more outlooks people bring to this, the better,” she observed.

Don’t forget lessons from history

There is a lot of fear surrounding AI and its parent, the so-called Fourth Industrial Revolution. The World Economic Forum has estimated that by 2022, 75 million jobs will have been displaced by automation. But, it also estimated that 133 million new jobs would be created as a result, and a large proportion of these would be roles based on human traits: customer service, sales and marketing, training and development, innovation management.

Former CTO Robinson said of this upheaval: “Every 20 or 30 years, jobs just go away, but what happens is, new jobs appear in their place. There is a lot of fear, it’s entirely natural to be concerned. But if you maintain a constant learning state and upskill where you can, your chances of moving into the new economy are better...This is not the first time the technology has changed. Learn about what has happened in the past. Appreciate the history. That’s the best way to understand what’s coming.”

Certain backgrounds will also give people an edge. Accenture’s Kannan, a quantum physicist by training before she went into tech consulting, shared that her educational background had actually been the entry point for her career in data science because she was deemed well-versed in math and physics. But she cautioned hopeful job-seekers to temper their expectations, particularly if changing careers. “If you have recently acquired a new data certification, even if you have 10 years of experience in another field, you can’t expect to go straight in as a senior data engineer!” she said.

Instead, she suggested that people keen on the field should develop an attitude of constant learning instead, because that gives them the ability to stay current in a very fast-moving environment.

“You need to want to learn new things,” she said. “That automatically will give you a sense of whether something is hyped or real. When I look for people, I look for that thirst for knowledge.”