In the past, artificial Intelligence was at first a technical problem. It was obvious that chess playing robots were not powerful enough and that biped robots have failed to walk upwards to stairs. With the advent of more modern algorithm it has become within reach to build much stronger AI systems which can do most of the desired tasks. It was succesfully demonstrated that AI can play Tetris, drive a car by it's own and cook a meal in a kitchen. From a technical perspective there is no longer a limitation visible. The engineers have figured out how to solve most of the problems, and the remaining tasks for example more general Artificial Intelligence will be solved by future scientists.
On the first look this sounds like a pathway to a wonderful future in which robot trucks are driving on the street and household robots are cleaning the dishes. Let us assume a world in which robots are demonstrated successfully for all domains, is this equal to build robots?
The difference between a robot demonstration a reall robot is that the later one is used for practical applications. Only if an autonomous car is sold to the public the technology is ready. The question is, when this will happen and does it make sense at all? At first let us understand what a common narrative idea. The idea is, that engineers are able to program autonomous cars and the consequence is that within the next 10 years these cars are marketed and sold to the general public. So we will see AI technology in the reality.
The first part of the statement is correct. Engineers have developed the technology and it was demonstrated many times. It is possible to control toy cars with a neural networks and larger cars as well. In a sense that the car takes car of the traffic light, obstacles on the road and maintains the correct speed. But, the second part of the narrative which is about introducing this working technology into the real world is under question.
How many examples for commercial robots are available in the last 50 years? Right not a single one. And the assumption is, that nothing will change in the next 50 years. Announcing a commercial robot is some sort of running gag but can't be transformed into a real product. At least for the last 10 years the common explanation was, that the hardware and software wasn't developed well enough. That means, it is too costly to build robots and if they are created from servo motors they are not able to solve tasks by themself. This prevents that a lab project is transformed into a valuable product.
With the advent of cheap microcontrollers like the raspberry pi and advanced software this bottleneck was solved. In theory there is no reason to delay the introduction of robots. Let us take a closer look into the situation today which is the year 2021. Not a single commercial robot can be bought worldwide. The only thing what is available are announcement of companies that they are planning to build robots in the future. The open question is will this happen or not?
Perhaps it makes sense to simplify the overall situation a bit and imagine some sort of robot in a sandbox. Suppose the idea is to build and sell a line following robot from scratch. The first thing to do is to build the hardware and then install a software on the device. The first stage for the prototype would be a working robot which is shown in a youtube video. Such a robot is able to navigate on a line and it can be used for many applications for example cargo transport. How many customers will buying such a robot? It is a rhetorical question because the amount is zero. The reason is that concrete customer has no purpose for such a robot. And perhaps this is the most surprising insight.
How can it be that a working line following robot, a chess playing software or an autonomous car doesn't fulfill the needs of the customer? The estimation is, that factories have a demand for a line following cargo robot and that private households need a self-driving car. And because of this assumption they will buy a robot if it is available at the market. But, is the assumption correct that there is a need for such technology?
What we can say for sure is, that in the past there was no need. The amount of kitchen robots and hospital robots sold to customers is known precisely, it is 0.0. that means, no one in the world have a demand for such a robot. The open question is why this should change in the future. Or let me explain it the other way arround. The implicit assumption is, that in the in next 10 years private households and large companies as well will buy lots of autonomous cars biped robots and kitchen robots. Will they?
Let us investigate what will happen if they don't. Not buying a robot means, that a robot is not delivered to the customer. And this is equal that the amount of sold units is low or even zero. The prediction is, that this kind of outlook is the more realistic one. That means, even if automative companies will announce self-driving cars they are not able to sell them because of missing customer demand.
Let us investiage which sort of product is highly demanded by customers. THis are classical computers like desktop pcs and smartphones as well. Also customers are buying lots of gadgets like remote controlled drones and fitness trackers. At the same time, the customers don't buy robots. How can it be that the customer has no need for automating existing process? The reason is, that a robot was never invented for such a task. Automating a process is done by mechanical machines like cars and with infrastructure like electric current. A robot was invented with the objective to increase the automation level further. And exactly for this additional step in technology a demand is not there.
Instead of focussing on existing companies who could develop and sell robot the more interesting platform is the kickstarter website. Kickstarter has made the introduction of new technology simpler. Some robots are shown on kickstarter. Most of them are companion robots which are talking to the owner and can do tricks like a dog. In contrast the amount of industrial robots or household robots is low. Why? Because of missing demand. The only example for a kickstarter robot from the domain of industrial appliation was a mini robot arm. But even this arm isn't sold for practical applications but according to the description it is a teaching tool to explain engineering students how to program robots. The chance is high that not a single real industrial or household robot was sold at kickstarter in the last 5 years.
And it is possible to make the argument more clearer. The chance is high that within the next 10 years nothing will change. That means no industrial robots at all are put on the kickstarter website and if a single example is available it won't find customers.
Let us stay for a while on the kickstarter platform because it allows to browse in the examples for robots. In the category robot most of the items are marketed with a certain plot. The idea is that somebody likes to know what robot and artificial intelligence is about and so he can bay a spider robot or a line following robot. The robot comes with an instructional manual and preprogrammed routines and then the human will have a lot of fun with the device. This is in short, the idea behind 99% of the robot at kickerstarter. But this use case is the opposite of a real robot. A real robot is marketed as a serious tool which helps to save time. Not the human but the robot should do a task. Unfortunately, kickstarter doesn't provide these robot and the plot isn't used to describe existing one. So it seems, that the customers have no need for such a plot. What the customer want's is to learn about robotics, but the customer doesn't need robot for practical applications.
Homecomputers
To grasp the difference between robots and homecomputers let us go back into the 1970s before the advent of homecomputers. At this time, the only way to become familiar with a home computer was to build such a machine by it's own. Sometimes a kit was sold in electronics store for amateurs. These kits were a great success because the result was that the customer was getting access to a home computer.
Since the early 1980s, homecomputers were sold as normal consumer products and the amount of people who have bought the technology is growing over the years. Nearly everybody has recognized how useful a computer is. The typical application of a home computer is to play games, write texts and program short programs. Because of this reason the customer are spending a lot of money for the technology and home computers have become a great success.
The assumption is, that for robots the situation is the opposite. First thing to mention is, that from a technical perspective it is possible since the 1980s to buy a robot by itself. That means, electronic experts do not need to buy commercial robots but they can tinkering the hardware in his garage. But nobody is doing so. Building a robot never has become a mass phenomena. Since a while, robot hardware is also sold in dedicated shops but the amount of sold unit remains low or even zero. It seems, that the customers have simply no reason to get it's own robot? And if they are buying a robot they are using the device never for practical applications but as a learning tool to understand what robotics is about.
Let us ask the question in a different way. If robots are so powerful why are no pictures available made of amateurs about self-created kitchen robots who are doing something useful? Technically it is possible to build and program a kitchen robot but it seems, that all the programmers in the world are not doing so. This stands in contrast to the homecomputer revolution In the 1980s it was a common situation that somebody has built a computer from scratch and used the machine for different applications.
In search for a kitchen robot
The most valuable source for searching working robot prototypes is youtube. There are two sort of kitchen robots available. First one are announcement of large companies that in the near future such robots will become available. Ii the meantime a prototype is shown which can in theory prepare a meal. The second example for kitchen robots are created by amateurs. Their DIY kitchen robot are always failed projects. That means, the robot is not able to grasp a bottle and this looks funny. So the amateur has captured the sceen in a video and other people like it. What is not available at youtube are working kitchen robots made by amateurs used in a meaning full way. The assumption is that such a situation is not needed and therefor nobody has created such videos.
The official description for a non working robot is a rube goldberg machine. The interesting fact is, that even a rube goldberg machine is working technically great it fails to fulfill the expectations. There is an example available about a rube goldberg machine who can make a pizza. On the first look this sounds like a practical applications of modern robotics. But similar to all rube godberg machines something is wrong. The result is that this machine remains the only one in the world and no other customer will buy such a machine.
In contrast, a working robot will produce a demand. A demand is equal that the robot is manufactored in higher quantities similar to the commodore 64 which was sold 17 million times. Let us go into the details. Somebody has created a single prototype for a pizza making robot. It is a not a real robot but only a rube goldberg machine The inventor has invested lots of hours to build and program the machine and then he makes a video about it. Other uses in the internet will watch the video and upvote it. Now something interesting will happen. After the robot was build, the inventor decides to disassemble the machine, because he needs the space in his house Also he will reuse the parts for the next project. So the project is lost for the world. And it is important to know, that even people who have seen the video won't rebuilt the project. Because it makes no sense to build the same non sense machine again. That means the total amount of pizza making rube goldberg machine was only 1 and it is not possible that the project will become a success in the future.
And this kind of lifecycle repeats over and over again. No matter if someone has built a line following robot, an autonomous car or any other AI powered device. The amount of copies is 1 but not more.
No comments:
Post a Comment