21st Century Robots

In about twenty years, everyone will be able to have their own robot just like they currently have a car or a smartphone, as new technologies, especially robotics, will contribute to shaping a world where artificial intelligence and robots will be involved in all areas of life. This is asserted by Daniela Rus, a professor of electrical engineering and computer science and director of the Department of Computer Science and Artificial Intelligence at the Massachusetts Institute of Technology at Cambridge University. She is not the only one who thinks this, and it is not science fiction.

The most famous futurist in the world, Ray Kurzweil, Google’s director responsible for new projects involving machine learning and language processing, author of the book ‘The Singularity is Near’, has predicted many changes driven by the development of new technology and has reportedly been correct in as many as 86 percent of cases.

Given the very rapid progress in robotics, it is hard to imagine which jobs smart machines will not perform in the near future, as robots that learn new skills every day have already been developed. The Japanese company Fanuc, the world’s largest manufacturer of industrial robots, has produced such robots, so it is certain that ‘mass slavery’ will mark the 21st century. Humans will enslave robots – or perhaps robots will enslave humans.

And while philosophers have not moved beyond the discussion of what being and time and ‘absolute nothing’ are for more than two thousand years, the fourth industrial revolution is underway, threatening the total dethronement of man. The economic power of the owners of new technology, artificial intelligence, and robots will be much greater than that of today’s arms and energy traders. But just as philosophers argue that time is relative, it is possible that changes during this century, in which very rapid technological advancement is expected, will mean an acceleration of not just a hundred but even 20,000 years. At least that is what Kurzweil claims, predicting that the power of computers will soon equal the power of all human brains on Earth and mimic the human brain, that we will experience virtual reality in about fifteen years with the help of implants, and that we will be able to ‘upload’ our minds into computers and thus remain digitally immortal.

Software will become the brain of the world, and as it becomes more perfect, technology will advance further, especially robotics, which has already proven its superiority in the application of robots in industrial processes. Fanuc engineers have programmed a robot that learns a specific process during an eight-hour shift, for example, taking items from one box and placing them in another. Experts refer to robot education as ‘deep learning’. During this, artificial intelligence uses a powerful microprocessor chip and algorithms that mimic neural networks and, with the help of large data sets, enable experiential learning.

The robot’s lesson of moving items from one box to another was recorded by a camera. The recordings showed that the robot ‘remembered’ what the object looked like every time it failed to perform the task accurately, which happened in only ten percent of cases.

Fanuc has invested more than seven million dollars in the development of the learning robot, and Shohei Hido, head of the research department at the company, stated that the robot learned the lesson during eight hours of work in the night shift. He also claims that learning can be accelerated if several robots work simultaneously and then ‘share’ the learned experience. For example, eight robots working together can learn in one hour what one learns in eight hours.

– Our project is focused on distributed learning. Can you imagine hundreds of factory robots sharing information? – asks Hido.

The method of distributed learning or cloud robotics is currently ‘in’ in robotics, and special attention is being paid to the development of robots that learn. Such robots, in addition to Fanuc, are also being developed by other companies, such as Google.

Although Ken Goldberg, a professor of robotics at the University of Berkeley in California, believes that Fanuc’s researchers, whose robots are in many factories around the world, can change the way robots are used in the near future with cloud robotics, he emphasizes that the ability of robots to learn is a significant challenge for robotics because it is very difficult to control their behavior; it is a much more complex process than, for example, recognizing objects in images.

– Deep learning has made great progress in pattern recognition, but the challenge for robotics is to do something beyond that – claims Goldberg.

However, there should be no bad students in the robotics school because by 2020 supercomputers will be up to 30 times more powerful than today’s, which will allow for the use of even more data and the creation of software of incredible power. It is predicted that the amount of data in the world by 2020 will be 44 zettabytes, or as many digital bits as there are stars in the universe, which will lead the development of artificial intelligence in an unprecedented direction. All of this will increase investments in robotics, in which the armies of the largest world powers and major companies like Google, Apple, Microsoft, and Facebook are already leading, investing billions of dollars in the development of robotics.

This could be a very lucrative business because robots have a significant advantage over humans, who can communicate with each other at only 10 bits per second; robots, on the other hand, can communicate over the internet 100 million times faster. Therefore, Gill Pratt, a roboticist and director of Toyota’s research institute, believes that robots in groups could learn ‘at lightning speed’.

However, Goldberg argues that we should not be overly optimistic and that predictions that artificial intelligence will surpass human intelligence are indeed exaggerated. His colleague Stuart Russell, a computer scientist at the University of Berkeley, agrees with him, stating that it is very difficult to predict algorithmic discoveries on which technological advancement depends.