Cybernetics means usually a man-machine interaction. A human driver operates a mechanical car, a human pilot navigates an airplane and a human worker uses a crane. All these techniques are working great and are the backbone of the economy. From the standpoint of robotics the challenge is to increase the automation level futher. That means to build a man-machine system which doesn't need a human operator anymore. The vision of Artificial Intelligence is to built cranes with humans, airplanes without humans and cars without humans. These systems have a higher productivity.
The problem is, that such a goal is unrealistic. It's technology not possible to built a cybernetic system without a human operator in the loop. Somebody may argue against the thesis by providing examples of robots who can pick&place objects by it's own. But these demonstrations are different from real tasks. There is not a single example available in which a robot has replaced a human worker.
The fact can be determined by a look into the literature from the past. Since the early robots in the 1960's companies have tried to use this technology for increasing the productivity. Many practical robots were build for serving a cup of tea, for cleaning the kitchen and for sorting apples. All of these projects have failed. The most obvious examples are robots at the production line of a car assembly line. On the first look, it's not very hard to increase the automation level, because the task is always the same and a robot is the natural choice for doing the work 24/7. The problem is, that all real car companies are working with humans at the assembly line. Not because they like it, but because robotics projects from the past were not successful.
From a formal perspective a man-machine system provides a certain automation level. A human operator who sits on a crane can provide a certain amount of work in an hour. A robotic crane has to proof that the productivity can become higher. And this is not possible. The hypothesis is, that it's technical possible to build classical man machine systems but it's not possible to build robotic systems which doesn't need a human operator anymore. That means, the automation level is fixed on the current level.
The question is, if robotics can be used in reality, what are all these robots projects about which were programmed by experts and amateurs? The answer is that Artificial Intelligence and robotics is a practical joke. It's funny to watch the results, but the technology can't be utilized for practical needs. A simple example for an interesting but non-sense software is an automated chess AI. Technically the software works great. The computer can play the game of chess much better than a human can do. That means, the engineers who have programmed the software are experts on the field. But this chess AI is useless for real applications. The reason is, that computer chess produces a new kind of game, unknown before.
And the same principle is working for a robotic crane as well. From a computer science perspective it's possible to program an automated crane. The system works in the simulation and with real objects as well. A planner is used who calculates the next movement of the robot. The problem is, that this crane is useless for a real construction site. It's only a joke technology, which is interesting to observe but has no practical applications.
The reader of this blogpost doesn't need to follow the hypothesis. He can try to search for a counter example, in which a robotic crane is used in the reality, for example to unload ships. But he won't find a single example. All the cranes in reality are controlled by human operator in the loop. It's not possible to remove the human and control the crane autonomously. This kind of insight is called by economist experts as the productivity paradox. It means, that the productivity in the industry is constant over the last 50 years and newly developed Artificial Intelligence won't change the situation.
Let us create some predictions for the future. One prediction is, that within 10 years it will become possible that humanoid robots are able to play a game of soccer. First robots who can do so are available in the Robocup challenge, and it's not very hard to predict, that within 10 years the robots will become much better in that game. Unfurtunately, none of these robots provide a value outside of the challenge. The robocup game was developed for the robots, but not to increase the productivity. The funny situation in 10 years will be, that on the one hand engineeres are able to build biped robots who can play better soccer than a human, and at the same time, the engineers struggle in using this technology for replacing human work with robots.
No comments:
Post a Comment