Unraveling the myths of robots
He left Google and the Bay Area to study in Sweden. Now Marin Šarić is taking on the daunting task of making a robot imitate human motions.
“What we are doing here is pioneering. It makes me feel that I’m working on something substantial,” Marin Šarić says.
In his otherwise neat office he is surrounded by loose robot body parts, heads, arms, and legs. The robot in the coffee room is learning how to take out a cup from the cupboard, preparing to someday be able to go to kitchen and take out a dish from the dishwasher without human supervision.
“People think we are at the point where robots will take over the world. It will take at least a hundred years before we can imitate a human brain and before it will reach industrial production,” says Marin, referring to the predictions of the world’s leading cognitive scientist Steven Pinker.
You can easily understand why. For example, take a part of research in computer vision where robots are designed to recognize human faces. There are thousands of papers published, some focusing just on how humans move their lips. It can take a full year’s study in leading-edge research to figure out just how to grasp a cup.
Craziest thing he ever done
“It’s a daunting task…there goes another PhD on such a small part of the human behavior,” he says laughing as he effortlessly takes a sip from his cup, a movement that’s very difficult for a robot to accomplish. Marin is going to help robots do motions that are really easy for us humans.
“Google already runs cars without drivers,” he says, making a passing reference to his former employer. After five years as a senior software engineer at Google and on a good career path, did he make the right decision to come here?
To embark on the master’s programme in Systems, Control and Robotics was the craziest thing he has ever done, he says: “After finishing, I only applied for a PhD here, no multiple options this time”.
As a 12-year-old in Croatia he got hold of some cool underground programs doing crazy animation tricks, some animating a Swedish flag. A couple of members of anonymous Swedish hacker groups he idolized as a young boy have now revealed themselves to be his colleagues.
“Actually the ‘programming’ of the hacker scene worked somehow because I’ve ended up in Sweden,” he says.
With a background in artificial intelligence that had little to do with systems and control he found the math quite intense. He had two serious talks with the programme director about quitting, but was persuaded to stick with it. For his master’s thesis, a young professor at the Centre of Autonomous Systems reached out a hand and offered him a subject he is pursuing as a doctoral thesis:
“It is absolutely fascinating, but daunting. There is no way I would have dared to look into this without her encouragement. In a field like this where everything is destined to fail you need a lot of support.”
As the robot starts focusing on what’s going on Marin is totally captivated. He cannot help but marvel at the human body.
“We can do such complicated things with such ease and it never even crosses our consciousness. The small things that make us work. But how much in reality is needed to just pay attention? It’s fascinating that in actually grabbing a cup you need so many different abstract clues from different areas,” he says. He talks about robots as if they are alive.
As a child he used to cheer the evil robots on TV. They were not after the destruction of human beings but simply wanted to be recognized for their feelings, he believed. Today Marin would be happy if he lived to see a robot as capable as a cat.
“My lifetime achievement would be to make robots as intelligent as a cat. They are not solving big cognitive problems, but they are very good in doing the basics.”
Text: Marie Androv
Edited by Brian Owens