ODENSE, Denmark — Not only can humanoid robots interact more naturally with people, but they can also help us consider important questions about ourselves, said Hiroshi Ishiguro during his keynote speech yesterday at RoboBusiness Europe 2016 here.
Ishiguro is a distinguished professor at Osaka University‘s Department of Systems Innovation, visiting director of ATR Hiroshi Ishiguro Laboratories, and research director of the JST ERATO ISHIGURO Symbiotic HRI Project. He is best known for his android replicas, which are among the most lifelike created so far.
The human-robot interaction researcher took the stage after greetings by Marianne Andersen, CEO of RoboBusiness Europe, and Anker Boye, Odense‘s mayor. Robotics Business Review is a sponsor of the conference, which is collocated with the Nordic UAS [Unmanned Aerial Systems] Event.
“This is my first time to give a regular talk in this country, about ‘Robots and Our Future Life.’ Usually, it’s about every creepy robot,” Ishiguro joked.
He noted that he has visited the country every year since 2011 because he is running collaborative projects with Danish universities and the Danish Technological Institute (DTI).
Ishiguro showed a slide of a fourth-generation copy of himself, which represents how he was when he turned 41. “Just as I’ve been improving the robot, I’ve been improving myself,” he said.
More seriously, Ishiguro asked, “What kind of knowledge can we get from a very humanlike robot?”
His Geminoid is a tele-operated android. “The goal is [for it to be] fully autonomous, but it needs to learn speech patterns,” he said. “The voice contains a lot of information, and we use a simple tele-operation system.”
Unpacking the self
Ishiguro noted that because he is in high demand as a lecturer, it’s easier to send his replica than to travel around the world more than he already does. Ishiguro recently gave a speech on short notice through his robot to the parliament of Chile, a 31-hour flight away.
“One characteristic of the fourth generation of my copy is that it’s portable,” Ishiguro said. “It breaks down into two pieces, plus the head, which is fragile — for carry on. A young professor can travel economy, cheaper than me, at half of business-class [airfare].”
“The robot can’t answer complicated questions, so I was tele-operating it from Japan,” he said. “But who has Ishiguro’s identity?”
“You’re interested in seeing the robot, not my body,” he noted. “I’m not creating this robot for fun but to deeply understand human nature.”
The first version of Ishiguro’s android received a lot of media attention in 2009. More recently, Ishiguro discussed “Autonomous Conversational Androids” at the South by Southwest Interactive Festival (SXSW) in Austin, Texas.
“SXSW started as a music festival, and it is now the biggest festival for interactive technologies and startups,” he explained. “Osaka University and NTT sponsored the presentation. NTT is the biggest telecom company in Japan, and it is interested in supporting speech recognition and chatbot technology.”
“People at SXSW enjoyed it very much,” Ishiguro said. “In USA Today, people said Pres. Obama was best presentation, then mine, from a technology standpoint.”
Why humanoid robots?
Ishiguro went on to describe his vision of a collaborative (but not necessarily industrial) future in which interactive robots and humans are together in daily situations such as a train station.
“Humans have brains that recognize humans, so the ideal interfaces for humans are humanoids,” he said. “For example, in Japan, there are already rice cookers, microwaves, and refrigerators that talk to users. They don’t need to read a manual” because speech is more natural.
Some technologies are “missing” or not yet mature for realizing the robot society, Ishiguro said. Better voice and image recognition, enabled by deep learning and cloud computing, is just beginning.
Just as personal computers eventually became cheap and reliable, so too do “we need to have a personal robot that is reliable,” he said. “The same thing happened with smartphones. If you just buy one, it would be very expensive,” but millions of devices provide economies of scale.
Ishiguro cited Pepper, which he called a “pretty good platform robot that functions almost the same as my robot. We spent $100,000 to $200,000 on mine; at just $2,000 for Pepper, that’s very cheap.”
“Aldebaran didn’t know the way of mass production, so it worked with SoftBank in China,” Ishiguro said. “The company is improving production lines, [and Pepper is] in Japan. They want to sell in China, the U.S., and Europe.”
Realizing the science
“We’re working on the basic science — autonomy, navigation, and manipulators,” Ishiguro said. “We propose interaction as the third important area, and human-robot interaction now has a conference and journals.
Ishiguro said he wants to develop humanlike robots based on advances in cognitive science. “We can verify hypotheses in cognitive science by using the robots, tightly coupled, but needed for development of any products.”
Virtual reality is also helping with tele-operated robots for different applications.
“We can simulate the sensation of touch, so an operator can accept the tele-opeated body as his own,” Ishiguro said. “With control by a brain-machine interface, EEG cap, SCR electrodes, imagine a handicapped person who can’t walk who could participate in society” with an android avatar.
Robert High, chief technology officer at IBM Watson, talked more about “embodied cognitive computing” later in the day.
Ishiguro told Robotics Business Review that he doesn’t know when the “singularity” predicted by Ray Kurzweil will be possible, “but it is our goal, to be totally integrated with machines.”
“Humans equal animals plus technology,” he said. “Consider evolution by genes or technologies — which is faster?”
Use cases for humanoid robots
Ishiguro argued that humanoid robots like Geminoid and Pepper are preferable to virtual assistants such as Jibo.
“When we got the PC, we used to use the virtual agent, computer graphics — but nobody is using them anymore,” he said. “Now we want tablet computers and smartphones.”
Ishiguro noted studies that have found that, thanks in part to stable voice recognition, language training can be a useful application. He elaborated on the “principles of conversation” in a later session.
“But agents need a humanlike presence; a 2-D screen isn’t enough,” he said. “It should be in the same physical space as us humans.”
Researchers have found that children and the elderly have readily accepted cute robots, leading to both commercial opportunities and potentially better care.
He observed that the Sota from Vstone Co. made a profit by taking pictures of children at kindergarten and uploading them to a website from which parents could buy them.
Commercial potential and immortality
“People didn’t need to accept the need for refrigerators or vacuum cleaners; they just needed them,” he said. “We need to do something to encourage our economies.”
“The smartphone is a combination of tablet computers and cellphones that’s easy to use,” Ishiguro said. “We need to understand how the human brain works, what humans need, then we can design better products.”
Ishiguro acknowledged that applications of autonomous androids might be a small market, with less-humanoid robots being more common. “Very humanlike robots will be like a Ferrari,” he said.
A popular comic storyteller in Japan was immortalized with one of Ishiguro’s lifelike replicas, granting eternal life through “human archiving,” he said. “Now that the national treasure has passed away, the robot is more popular than the original — which presence is stronger?”
‘Beauty is truth, truth beauty’
Ishiguro said that replicas of fashion models, “idol singers,” and a TV talk show host show that “beautiful humans are not humanlike, but we want to be beautiful.”
“It doesn’t need to go to the toilet; the android is the ideal human,” he said to laughter in the audience. Robots could represent idealized forms just as classical statues represented gods or the Buddha.
About three years ago in Odense, an “android theater” performance showed that a robot “could represent human hearts and minds, but it is a simple computer program,” Ishiguro said. “It sounds like a subjective phenomenon, but the audience said it felt a human presence, even though it was just a simulation.”
Such identification can help humanity reflect on itself, be used in therapy, and sell products. Ishiguro pointed to cultural resistance to asking for directions or help in a store, but with robots, “men never hesitate — they ask many questions.”
“You can sell a lot of products, more than humans could,” he said. Robots can flatter guests, Ishiguro said, “and robots never tell a lie. In Japan, people never check change from vending machines, but they do from [human] shopkeepers.”
Journalist Antje Gerd Poulsen asked Ishiguro if he would consider robots as a new species. “Yes, it’s a marker of society. Robots are not human, but they’re not machines or monkeys. They’re in between,” he said.
“In daily situations, if there is no way to tell which is which, we’ll accept robots as close to human in terms of rights,” he said. “People want to be equal.”
Leaving room for the imagination
While Ishiguro is known for working on very humanlike robots, he has also found that less-humanoid designs can also be useful.
“One of the most important findings of my research is that recognition based on observation is how to overcome the ‘uncanny valley,'” he said. “But recognition based on imagination — with a neutral, roughly humanoid appearance and humanlike voice, the robot activates the user’s imagination.”
Tests in nursing homes found that the elderly were less nervous, and people with dementia were less nervous and could speak more normally around more abstract androids.
“We can minimize modalities to encourage imagination,” Ishiguro said. “What are the minimum media conditions to sense a humanlike presence?”
The Hugvie, a stuffed doll with a smartphone holder, reduced stress. “We tested saliva and blood and found reduced cortisol significantly vs. just a voice on a cellphone,” Ishiguro said.
“Two modalities is minimum — voice, tactile, smell, appearance — we should consider different combinations,” he added. Rowdy schoolchildren were given Hugvies playing their parents’ voices, and they became more orderly.
As Ishiguro and his teams continue to develop more human-friendly autonomous robots, they are looking to “intentional design” — inference of desire and simulating emotion.
“What is the human heart?” he asked. “There is no way to verify imagination, understanding.”
An audience member from Spain asked about how to promote imagination with prior experiences.
“We are thinking of how to manipulate imaginations through cognitive research,” Ishiguro responded. “Hypnosis is very interesting.”
“Everything exists in society, story controllers,” Ishiguro said. “Humans are story machines.”