Media partner Synced attended the AI and Society Symposium held October 10-11th in Tokyo, where speakers from industry and academia exchanged ideas for practical applications and possible future developments of AI technologies. The symposium sought to bridge the gap between cutting-edge research in academia, state-of-the-art AI applications in business, and safety concerns about AI development; and to stimulate worldwide discussion on social impacts arising from new AI technologies.
Speakers covered a wide spectrum of backgrounds, from researchers on machine learning and artificial general intelligence (AGI) to philosophers on ethnic and legal issues; from artificial consciousness, development entrepreneurs to venture investors; from computer scientists investigating artificial life evolution to artists working with AI.
There were five keynote speakers: Prof. Hiroaki Kitano from Sony Computer Science Lab (CSL), Prof. Oren Etzioni from the Allen Institute for Artificial Intelligence, Dr. Akira Sakakibara from Microsoft Japan, Marek Rosa from GoodAI and Prof. Hod Lipson from Columbia University.
Prof. Kitano shared his thoughts on the Nobel Turing Challenge. As a system biology scientist, he found that knowledge discovery is widely applied in biology research, which is typically a non-linear, complex process difficult to execute using human intelligence. He, therefore, has developed an online AI platform called Garuda, which hybridizes human and machine approaches on system biology research. He argued that the complexity of living systems entails the study of biology in multiple dimensions of time and space. Motivated by this high-dimension complexity, the Garuda platform gathers data and gadgets from biology researchers around the world and captures the complexity at a system level.
Prof. Kitano also advanced an ambitious proposal: Instead of evaluating a machine’s intelligence using the Turing Test, a new intelligence challenge would be building a computer system able to make scientific discoveries by using AI systems’ knowledge discovery ability. This would give us a deeper insight into and perhaps redefine the scientific process, while the resulting discoveries could even qualify for a Nobel Prize in physiology or medicine. The authentic Nobel prize can only be awarded to humans at present, might Kitano’s challenge change that? This challenge may serve as an alternative to the Turing test for the AI intelligence, like the Nobel Prize for the human intelligence.
Dr. Oren Etzioni spoke on the topic “AI good or evil?” Dr. Etzioni is the CEO and Chief Computer Scientist of the well-known independent research institute AI2. He asserted that the depiction of an “evil AI” is far from the current reality of today’s AI technology, and portrays AI technologies in a negative way. He defined a harmful intelligence as an autonomous machine. For example, a computer virus is usually autonomous but not intelligent enough to harm humans on a great scale, while game-playing AI AlphaGo is intelligent but not autonomous because Go is the only task it’s good at. He said his 6-year-old son “is more autonomous than any AI system. He can make his choices, he can cross the street, he can explain himself, he can understand English when he wants to. That’s the difference.” All in all, Etzioni argued that despite the hype and headlines of AI doomsday scenarios, it is already positively impacting the real world and can save thousands of lives in the coming years. “AI is a technology, and the choice is ours.”
Prof. Hop Lipson from Columbia University predicted six development waves in an AI timeline: Established symbolic computing, the current application of predictive analytics to cognitive computing, creative machines, physical embodiment machines, and finally sentient computing. We have developed AI from rule-based to data-driven, and with the availability of big data (yes, data is precious) and the development of machine learning techniques, machines have learned to distinguish dogs and cats in the last five years. Dr. Lipson expects the next AI breakthrough will be driven by cloud computing. Having worked on self-aware and self-replicating robots which challenge conventional views of robotics, he predicted that physical embodiment of AI will happen in parallel with development in material sciences such as the artificial muscle recently invented at MIT. Finally, he said he would not be surprised to see sentient robots/machines realized in the next few decades.
In other keynote talks, Dr. Sakakibara covered the latest trends in AI and the “Democratizing AI” strategy, and Marek Rosa presented his roadmap to building general AIs based on a hierarchical skill development.
There were quite a few other interesting talks. Prof. Kenneth Stanley from Uber and Central Florida University discussed his attempts at designing an open-ended learning system mimicking the learning of a biological system; Youichiro Miyake from Square Enix talked about how his team designed the intelligent gaming system in Final Fantasy XV; Prof. Shun’ichi Amari proposed his hypothesis and concerns about the problem of artificial consciousness and free will from a mathematician’s perspective; and Prof. Selmer Bringsjord showed how his team is attempting to solve the artificial general moral intelligence (AGMI) problem with logical formulations.
Journalist: Joni Chung | Editor: Hao Wang, Michael Sarazen