Top of software development department of GE talks about "How to engage with human beings evolving with artificial intelligence and machine learning"

With software technology such as machine learning, artificial intelligence (AI) technology is evolving at a learning pace that is far from comparable with the past, Colin · Paris, vice president of General Electric (GE) software research department In an interview with PCMag, I am clarifying my view on how human work changes with the evolution of AI.

Skynet Is Actually Real (But Hopefully Will not Try to Destroy Us) - YouTube


Paris, who mainly leads the project to analyze the big data, analyzes the data accumulated in the jet engine business that has been going on for over 40 years, the wind power generation business for more than 20 years, etc. and when maintenance is the best I am trying to find the maximum efficiency and safety at the lowest cost. Among them, they are using machine learning techniques to find specific patterns from data.

In addition to machine learning in GE "Digital Twin"We introduce a technique called. Digital Twin utilizes big data as input data in a method that real-time input of a large amount of data obtained from a large number of sensors, reproduces the real world event of the manufacturing site in the digital world and increases the accuracy of the simulation It is a simulation. By using the digital twin, it seems that it is possible to predict the timing of maintenance necessary for jet aircraft, the life of the aircraft itself, the minimum fuel and cost required for flight, the amount of exhaust gas, etc. with high accuracy.


In response to the question "When do you think that AI will become a commercial mainstream?" Mr. Paris replied, "Now it is just the main stream." Google predicts what the user expects from search phrases, makes effective advertisements, recommends products that suit Amazon preferences, etc. It is a technology that leveraged AI, already AI is a real service That is why it is introduced in.

For many people who are feeling suspicion that AI will replace humans, "The work of ascertaining rust (rust) in oil rigs that dig in oil from the bottom of the ocean is from the ground 60 I was hanging people with a rope from a huge scaffold that was assembled at a height of meters.It is a dangerous inspection work that can be overwhelmed by strong winds but now drone is substituting.This is a camera "There is no need for humans to do dangerous work anymore by analyzing images to see if they are rusty or damaged," he says. "Work like 3K" which is dangerous, dirty and tight, and monotonous It is requested that AI substitute for boring work etc, says Mr. Paris that there is a field where machines replace humans.


Furthermore, since it is not possible to predict the arrival of new technologies or services in the future, it is difficult to accurately judge the threat of alternatives to AI, but machines can not substitute, for example, in areas where communication with people is required Mr. Paris thinks that machines are required to assist human beings in these fields in order to make human beings more comfortable in these fields, he says that forms of working with human beings and machines are sought.

In response to the question "Do machine learning safe automatic driving car?", Mr. Paris has a positive opinion, "I think so". According to Mr. Paris, Tesla's EV collects data for each mileage of 1.3 billion miles (about 2.1 billion kilometers) in 2016 from each driver. From the fact that when a boy throws into the road a ball or dog tries to cross the front of the car, the human beings stop the car by stepping on the brakes EV learns to stop in this situation in order to avoid danger It seems to be said that the safety of automatic driving cars is increasing at learning speeds that are not comparable to humans by collecting large amounts of data.

For the question "What is a thing you can not do with a machine by human beings?", Mr. Paris explained as an example of an automatic driving car. Google's automated driving car who discovered "a wheelchair walking a dog" was able to distinguish a dog, but he said that a wheelchair person could not distinguish something. Moreover, the automatic driving car seems to be unable to understand the hand sign that the cyclist gives out, and he says that there is still a need for human beings to support the machine even in such a scene.


In any case, Mr. Paris thinks that learning pace is accelerating thanks to machine learning at 10 times faster than before, and the evolution that will occur in the next 20 years is the past 200 years and how people I think that the machine may reach the area of ​​thinking to think about.