Event Report: C&C User Forum & iEXPO 2017

5th NEC Future Creation Forum

Session 2 Part (1)
The impact of advancing AI on future society

Presentation 1: Deep learning will change society

Yutaka Matsuo, Project Assistant Professor, Graduate School of Engineering, University of Tokyo

Deep learning technology has made rapid progress in the field of still picture and video content recognition. Robots can perform more complex tasks better by combining deep learning image processing techniques with reinforcement learning.

Put simply, these changes involve the “birth of sight.” The Cambrian Explosion suggests that the acquisition of sight enabled living organisms to develop diverse living strategies, and species to diversify. The same thing could happen with machines and robots. If we combine image sensors, or retinas, with deep learning techniques for managing the underlying visual cortex, we could create the first “seeing” robots. This development would greatly impact industries such as agriculture, construction and food processing, which have been difficult to automate and mechanize.

Having said that, deep learning has far inferior cognizance than living things because living things are able to create loops that apply cognizance to constantly changing patterns and correct behavior accordingly. Humans can also apply codes to create conscious thought. The future development of deep learning research should fuel rapid progress in future artificial intelligence technology, and help acquire early understanding of language, science and laws of physics.

Once that happens, humans would be able to focus on jobs that promote communication between people, set goals and judge value. Researching intelligence forces us to consider who and what people are, and analyze human qualities other than intelligence.

Portrait of Yutaka Matsuo

Project Associate Professor, Graduate School of Engineering, University of Tokyo

Yutaka Matsuo

Presentation 2: Technological progress and people

Yoshiharu Habu, Professional shogi player

Software opponents are getting stronger in Japanese shogi chess and other board games. Up until now, robots have performed brilliantly in certain, limited areas, but recently AI is able to apply moves, ideas and complex judgements that humans cannot hope to understand.

What is changing so dramatically here is not the ability to collect large amounts of data and make multiple computations, but the ability to make accurate evaluations. In its simple form, evaluation is about determining a good or bad move, but we also need to be able to agilely adjust that evaluation. Technological advances mean software is much more accurate and displays astounding speed. Shogi is already utilizing that power to help analyze and research improvements.

People have clear blind spots and blind angles. Getting software to pinpoint those blind spots helps us expand our thoughts and creativity. Humans are the sole high-level intelligence species, but combining and comparing human intelligence with a different form of intelligence such as AI is already creating new possibilities.

One point that concerns me is that the process of deep learning feels like it is occurring in a black box and we don’t really know what is happening. Theoretical physicists such as Dr. Michio Kaku understands the more in-depth concepts such as the 9th or 11th dimension. Contemplating those complex concepts could be key to understanding just how AI or shogi software works, and, as technology develops, understanding more about how the human brain operates.

Portrait of Yoshiharu Habu

Professional shogi player

Yoshiharu Habu

Panel Discussion: Should AI be feared?

Chiaki Hayashi, CEO of Loftworks and moderator of the NEC Forums asks the inevitable question: “We hear much that is positive about AI development, but should humans be scared of a world they cannot fully control or comprehend?”

Scene photo from panel discussion

“In shogi, software is getting stronger and is often more precise than humans, but ultimately it makes mistakes. Relying absolutely on these new creations would be extremely dangerous. They are extremely convenient and offer significant social benefits, but they require human support and intervention,” says Habu.

Kevin Kelly, Founding and Executive Editor of WIRED magazine, is a strong proponent of developing complementary relationships between humans and AI. “After losing to the most advanced AI, Deep Blue, world chess master Gary Kasparov created a human-AI team that became the best player on the planet. Maybe this could work in shogi or go. Humans and deep neural nets both have blind spots. You can fool a deep-learning algorithm by making a slight change to a picture of a cat, which we would still recognize as a cat, but AI would now think it something completely different. AI can help humans, and humans can help AI recognize and compensate for their own blind spots.”

Commentator Chiki Ogiue says fear should not be directed at AI itself but at the inertia of constricting technological advances. “I don’t see it as a battle between AI and humans. When letters were invented, Socrates feared “people would become lazy and lose their intelligence,” but we used words to acquire wider knowledge and enhance our intelligence. We should be positive about developing future complementary human-AI relationships, because social inertia and clinging to the status quo will restrict our relationship with technology, and reduce our future scope.”

Ryojin Shionuma, chief priest of the Jigenji Temple advocates a balance between establishment and anti-establishment forces. “I believed computers and AI had advanced enough already but these discussions have helped me view AI as a useful tool for promoting coexistence in our universe, planet, countries and cultures, with humans in control to help create a better world.”

Portrait of Dr. Michio Kaku

Professor of Theoretical Physics, City College of New York

Dr. Michio Kaku

Dr. Michio Kaku, theoretical physicist and professor at the City College of New York puts AI’s current capacity into perspective. “Can AI replace human beings? Equating robots with animals, the most advanced would be a cockroach. By the end of the century, if robots may develop self-awareness and achieve the intelligence of a monkey, they may become more of a danger. However, robots cannot do three key things. They can’t interact with humans. Even the most advanced robot today is no better than a tape recorder. They can’t do simple tasks such as put out the trash, or anything that requires some skill. Second, they lack common sense. A robot doesn’t know that water is wet, or that a mother is older than a daughter. Third is pattern recognition. Progress has been impressive but if you put the most advanced robot on a street, it couldn’t cope with the myriad of patterns and would likely get run over by a car. The time for concern about robot intelligence will be around the turn of the century, when self-aware robots will probably be able to fix a leak or a faulty toilet, or design a garden. But that is nearly 100 years away.”

How will AI change society?

What changes will our everyday society undergo over the next 10-20 years as humans seek to coexist with AI?

“We shouldn’t think of imposed change, but humans controlling computers and robots,” says Shionuma. “We are human, so we sometimes feel pain, sweat, and shed tears, but this enhances our personalities. Whatever change occurs, it will be led by humans, and that elective change will impact the way our relationship develops with robots.”

Scene photo from panel discussion

Habu believes there are two possible ways that AI could benefit education techniques and society. “We could either use a math-based approach to analyze vast volumes of big data, or use personal data from the instant a person is born to determine the best way to educate and learn. Also remember, many people achieve incredible results far away from the digital world, and that will become increasingly important as the digital world advances.”

“What will change will be social norms, but we must change them prudently to ensure greater happiness,” says Oguie of cultural change. “In the past, society was starved of resources, so people had to conform to certain set rules to avoid confusion. With multiple information processing possibilities, people will be able to act in different ways without overwhelming social capacity. We can create a social structure in which convergence helps minimize the unhappy aspects of live, and divergence maximizes individual happiness. In the past, “happiness models” were bound by standards for how a family or a person should be. As technology advances, we will be able to focus on “how not to be unhappy,” and create more diverse standards and norms.”

“One of the virtues of technology is that it unearths new needs and desires we weren’t aware of,” says Kelly. “Every day, Google answers one billion questions. 20 years ago, those questions weren’t asked or answered, so Google, an AI, has changed our behavior. Predictions is another area where AI will likely change our behavior. Predictions used to be very expensive, but AI makes predictions much cheaper, or even free. AI predictions will start to feature more prominently in our lives, bringing benefits we were not yet aware of.”

When discussing how humans and AI coexist, Katsumi Emura, NEC Chief Technology Officer, believes the chief debate is not artificial intelligence but augmented intelligence.

“We have to imagine how to use AI to enhance human intelligence. Whether we are considering diverse or concentrated education, as technology advances, we will need to make conscious and responsible decisions on how to use that technology and how we change society, because that will become our society of the future.”

Moderator Chiaki Hayashi concludes that, “Advancing technology is not enough to make us happy, we have to design and mold it to create the future we desire.”

Scene photo from panel discussion

> To be continued: Panel Discussion 2