Boston Dynamics
According to a report from Bloomberg Technology, Google (NASDAQ: GOOGL) is looking to dump Boston Dynamics, a robotics group at the forefront of the company's artificial intelligence (AI) efforts.However, in March, Google's AI sector appeared to have taken a step-forward, having won a board-game tournament against against one of the world's top players (robot versus human).
Specifically, Google's AlphaGo programme trumped Lee Sedol -- one of the world's best Go players-- in a best of five Go competition in South Korea. Sedol was astonished. In fact, most experts figured the 3000 year old, ancient Chinese game (Go) was far too complex for computers to master. But AlphaGo operated with masterful touch, displaying unorthodox moves that initially perplexed observers but ultimately paved the way for victory.
Contrary to the success of AlphaGo, the direction Google has gone with Boston Dynamics is far less victorious.
Note that modern industrial robots are "dumb" machines whose operations rely upon pre-programmed routines. They do not actively respond to their environment, but rather, follow limited instructions. Furthermore, these "dumb" machines are housed in big metal cages (meant to inhibit people from walking into their zone of movement), making them clumsy and the antipathy of nimble.
Boston Dynamics, under Google's guidance, looked to break free from these rigid confines and develop a product capable of performing a much broader range of tasks. Unfortunately, such a task manifests a range of unsolved problems that necessitate intense fundamental research. At a meeting held in November, Google's robotics leaders and AI executives discussed the viability of teaching robots to respond to their outside environment and do physical tasks. It was suggested that Boston Dynamics needs to collaborate with other Google teams.
Teaching robots how to do physical tasks on command requires advanced hydraulics. While the robot industry is excited about "smart machines," even the most advanced companies, like Google, are still trying to figure out the necessary hydraulics to build complete machines.
Thus far, Boston Dynamics' robots move in a realistic manner, but are hardly capable of performing with full autonomy. In essence, a robot cannot walk itself, as would a dog. It does not have a mind of its own-not yet at least. For instance, indoors, Boston Dynamics' robots can stack boxes autonomously, but only if a human sets everything up and tells the robot when to start. Herein lies Google's problem. It's tough to give a machine intent. Programmers can simulate intent in a virtual setting, but translating those movements to the real-world is extremely difficult.
Within these difficulties there exists a research dichotomy: "simulate an environment and orbit with software and hope the results are accurate enough that you can load it into a machine and watch it walk." Or, "skip the simulation and tinker directly on a robot, hoping you can learn things from the real world." Problem is, the latter is a very slow process, while the former is conducted with rapidity.
Simulation can be used to test out different scenarios and then analyze how the robots react. Meanwhile, real-world trials provide Google with data and problems that virtual simulations do not provide. But even simulation is difficult, as robots are rifled by structural complexity. Robots have arms, necks, knee joints, and fingers. Coding for every aspect of a robot's structure is time-consuming. Testing out the different ways a robot moves in flesh-and-blood reality takes even longer.
In reality, Boston Dynamics' robots need technology that does not exist. Such technology would allow researchers to handle the diversity and complexity of a robot's environment, further enabling them to respond to the randomness of everyday tasks. The software required to control robots and give them full autonomy is a research problem being worked on by universities and tech firms around the world.
Conflict of interest
Executives at Alphabet Inc -- Google's parent-- are laser-focused on revenue. In fact, the main reason Google is putting Boston Dynamics up for sale is because of revenue -- execs realize that the business unit isn't likely to produce a marketable product in the next few years. No marketable product means nothing to sell, which means no revenue. For a corporation looking to boost its stock value, revenue is crucial. But Boston Dynamics isn't concerned with revenue -- they're concerned with R&D, and ultimately, the future. As such, they're up for sale.
In terms of potential buyers, both Toyota Motor Corp (NYSE: TM ) and Amazon.com Inc (NASDAQ: AMZN) have expressed interest.
Speaking of Toyota, a main competitor-- Honda (NYSE: HMC)-- has developed its own humanoid robot named ASIMO. ASIMO stands for Advanced Step in Innovative Mobility, and is purportedly the world's most advanced humanoid robot. Honda created ASIMO as a helper-- be it helping around the house, or assisting an elderly/disabled person confined to a bed or wheelchair. In response to simple voice commands, ASIMO can comprehend and carry out tasks, and even hold conversations in Japanese, English, and Chinese.
By 2020, Honda aims to use ASIMO robots at train stations and large commercial facilities as tour guides.
Human-robot interaction
The more this research problem is solved, the closer we approach a future in which humans and superintelligent computers coexist, work together, and maybe even develop relationships. Both Hanson Robotics and Hiroshi Ishiguro Laboratories seek to make this far-fetched future a reality.
In light of these efforts, Dr. David Hanson has developed an android name Sophia. Sophia's skin is made from patented silicone, and she's capable of emulating more than 62 facial expression. She uses a cameras inside her eyes (alongside computer algorithms) for facial recognition. In addition, she can process speech, chat with humans, and actually get smarter over time. The end goal for Dr. Hanson is to develop a robot as conscious, creative, and capable as a human.
If robot development reaches these heights, they'll be of service to health care, therapy, education, and customer service applications. Furthermore, the most human-looking robots will be valuable receptionists, museum tour guides, and language tutors.
Hanson's hope is that by the time he's finished, robots will be indistinguishable from humans, capable of walking, playing, teaching, and forming relationships with humans. Hanson will announce pricing and availability of his humanlike robots later this year.
Meanwhile, Dr. Hiroshi Ishiguro has created a robot named The Geminoid, which dons a plastic skull, metal skeleton, and silicon skin. The Geminoid is controlled by an external computer, but looks like a normal human in remarkable ways. In fact, 80% of the people greeted by the Geminoid thought the android was a real human. Considering the hefty $100,000 price-tag, Dr. Hiroshi doubts the average household will indulge.
Much more accessible, however, are his smaller, cuter robots named CommU. The CommU is a communicative robot that uses voice recognition technology and AI to simulate conversations. They're useful for tutoring and elder care. For instance, in Japan, elder care is in high-demand due to their rapidly aging population. Robotic caregivers keep the elderly occupied, particularly with people who have dementia. In addition, these robots can offer to play games, and even dance with the elderly, which keeps their mind active.
Undoubtedly, humanoid robots are still in their development stage. But perhaps not for long. Based on the progress made by Honda and Boston Dynamics, it looks like 2020 is a realistic time-frame for the deployment of humanoid robots.
NOTE: Currently, Boston Dynamics is still up for sale.