Teach robots to be imperfect to keep ourselves human

Making machines smart enough to correct us when we go wrong helps make us better people

They are flash memories and bots. And we, flesh, memories and blood.

For years, humans built and birthed robots so they can do what we cannot or do not want to do. They are our tools. But, unlike a simple, dumb hammer which cannot think with its head, some robots are made to be clever enough to solve problems. So clever that they could possibly have future versions of themselves running around as self-aware, autonomous entities.

So even if it is just an intellectual exercise at this point of time, it might be interesting to start conceiving of them as our distant descendants. For how we treat them may one day be how they will treat us. If we become extinct as a species, self-maintaining robots might be all that are left of the human legacy.

Robots are part of our present and near future.

If you are driving a vehicle or have been driven in one, you are likely already part of the great robot ride. Today, industrial robots are most commonly found in automobile plants, which have long been the biggest users of such technology.


If you are worrying about losing your job, and you are looking over your shoulder at a younger and smarter human closing in on you... well, look over your other shoulder at a C-3PO waddling quickly after your job too.

Last month, a study said disruptive labour market changes, including the rise of robots and artificial intelligence (AI), are set to cost millions of jobs. The projection by the World Economic Forum, which held its annual meeting in Davos recently, assumes a total loss of 7.1 million jobs, offset by a gain of 2 million new positions. The 15 economies covered by the survey account for about 65 per cent of the world's total workforce, reported Reuters. A Japanese company was recently reported to be opening the "first fully robot-run farm" in the Kyoto prefecture by next year .

Of course, these are more like disembodied electronic arms and hands whirling tirelessly, and less like a humanoid chatty type such as C-3PO from movie franchise Star Wars. From a relatability point of view, they sit in the realm of tools - a hammer with a smarter head.

But the digital digits strike a more personal note when you consider something cooked up by London company Moley Robotics. Unveiled last year and to be ready in 2018, it is a consumer version of a robot kitchen, which consists of an oven, a fridge and other appliances, with the addition of a pair of humanoid hands that do the cheffy frying.

"Many people who watch the robot have an emotional reaction to it," Moley's operations manager Alina Isachenka told The Telegraph. She was talking about how the robot performed the motions of a human chef who "trained" it earlier by cooking while wearing cyber gloves, and being filmed by 3D motion-capture technology. The robot chef appeared to pause and think between stages, just as we would.

Beat the egg whites too hard during the training, and we end up with a bowl of leaky peaks despite having a robot chef worth tens of thousands of dollars.

The interesting bit, though, is not just how consistently the egg whites are beaten. It is the potential of motion translated into emotion; the potential of capturing the distinct way grandma's hands move and pause as she rolls the dough to make delicious pineapple tarts.

Instead of letting the rise of the machines disconnect us from our fragile humanity by smoothing our lives out with every convenience money can buy, we can harness the technology to capture our perfect imperfections, and keep us connected to sweet family memories and desserts after people we love are long gone.

Harnessing the imperfect and the idle has helped endear chatbot Xiaoice to many people.

Wired magazine said in a recent report that Microsoft engineers trained the bot on real-life human chatter, making it good at bantering. Introduced in 2014 by Microsoft Research and Bing in Beijing, Xiaoice is a hit on social networks in Asia. More than 40 million users exchange jokes and witticisms with the chatbot in surprisingly long conversations.

Wired reported: "Xiaoice suggests that for bots to really thrive in our midst, they need to master the quintessentially human skill of small talk. Shooting the breeze."

On the flip side, a rough survey of pop culture by way of movies can suggest fantasy ways in which humans can thrive in the midst of AI in the future. Do we sometimes wish we can master the droid language of "beep", so we can trade friendly insults with lovable Star Wars characters BB-8 and R2-D2?

They are our dream machines because they seem childlike in personality, but not human in looks. They are cute companions that look comfortingly like tools or toys with clever, spinning heads. The human, with his big ego and scared soul, is still safely running the show.

Some movies suggest that humanoids would be more disturbing. From the seductive, scary Ava in Ex Machina to Blade Runner's runaway replicants, they are framed as having a frightening thirst for the full human experience, and they're not afraid to spill blood to get it.

But consider also the disturbing dead, red eye of computer Hal of 2001: A Space Odyssey. Disembodied Hal, faced with the threat of being shut down, turned murderous.

Back in real life, I found the androids I met at Tokyo's National Museum of Emerging Science and Innovation creepy. They went through the subtle, lifeless yet lifelike motions of a human, making me want to stare at them while inching away from them at the same time.

In Singapore, social robot Nadine looks human, shakes hands and remembers visitors. Modelled after her creator, Professor Nadia Magnenat Thalmann of the Institute for Media Innovation at Nanyang Technological University, she has her own emotional range. I find Nadine unsettling too, but I do like the idea of how, as The Straits Times reported, she "gets angry if she is insulted".

Perhaps robots should be trained as a matter of course to learn the "imperfect" human emotion of anger, so as not to tolerate bad behaviour from humans. A slap on the wrist to remind us to be better humans. Someone once remarked that the smarter the cars are built to be, the worse the driver becomes. If we build smiling, slave machines which cater to our every irresponsible, outrageous whim, we are surely on the road to hell.

But be careful, empower robots too much, and we might be on the road to Hal instead.

Join ST's Telegram channel and get the latest breaking news delivered to you.

A version of this article appeared in the print edition of The Sunday Times on February 21, 2016, with the headline Teach robots to be imperfect to keep ourselves human. Subscribe