Robots that defeat humans in a game of chess or computers that we can talk to – science has been trying for many years to artificially recreate the complex human mind. How far have they got with this?
How intelligent is artificial intelligence?
The research area “Artificial Intelligence” (AI) tries to simulate human perception and human actions by means of machines. What once started as the science of computer programming has more and more developed into the study of human thinking.
For after decades of research, the impossibility of creating a “thinking” machine has been recognized without having first explored and understood human thinking itself. This is why there is sometimes a large overlap between AI research and neurology or psychology.
Until today, it has not even been possible to reconstruct human intellectual achievements as a whole with machines. A major obstacle is speech processing. Even the execution of the simplest commands is a highly complex process for a machine.
Research is therefore increasingly concentrating on individual sub-areas, among other things with the aim of making work easier. This requires a constant exchange between scientists from different disciplines (cognitive science, psychology, neurology, philosophy and linguistics).
Many scientists distinguish between strong and weak AI. Weak AI only covers partial areas of intelligence. It is usually based on methods from mathematics and computer science and is used, for example, in navigation systems and speech recognition.
Strong AI involves logical thinking, planning, communication, and making independent, complex decisions. Many researchers doubt that it will ever exist.
And even if it did, many ethical questions would arise: Which decisions can be left to an artificial intelligence that has no morals and no awareness of right, wrong and above all of nuances?
When does a computer pass the Turing test?
The question of when a machine is considered intelligent has been driving AI research for decades. A measuring tool that is generally accepted is the so-called Turing test.
It was developed in 1950 by the British mathematician Alan Turing: A human being communicates over a long period of time in parallel with another human being and a machine without visual or auditory contact – for example via a chat program.
Man and machine try to convince the tester that they are thinking people. If after the conversation the tester cannot say with certainty which of the conversation partners is a human and which is a machine, the machine has passed the test and can be considered intelligent.
- In 1991, the US sociologist Hugh G. Loebner offered a prize of 100,000 dollars for the computer program that passes the Turing test and tricks a jury of experts.
- In 2014 there were reports that the Russian software “Eugene Goostman” had passed the test. However, doubts were raised afterwards about the method and the test arrangement.
- By 2020, no one had received the award, and the majority of AI researchers assume that this will not happen in the foreseeable future either. A robot solves a magic cube.
Will there ever be a robot that can think?
The fields of application of artificial intelligence are extremely diverse. Often we are not even aware of them. Their use is most successful in small sub-areas such as medicine: robots perform certain surgical procedures – for example, in the range of a thousandth of a millimeter – much more precisely than a surgeon.
In production lines, especially in the automotive industry, robots replace a myriad of human hand movements. Particularly when it comes to health-endangering, accident-prone tasks, such as painting or welding, it is impossible to imagine life without robot arms such as those used by General Motors back in the 1960s.
A classic application area for artificial intelligence are games, especially board games like checkers and chess. Programmable and adaptive toys, mini-robots and computer programs have long since conquered the children’s room.
The legendary Tamagotchi is already old hat, but other artificial companions such as robot dogs, talking dinos or dolls, with which one can communicate through simple gestures or language and which perform certain tasks, are pushing onto the market.
World chess champion Garry Kasparov 1997 playing against the computer “Deep Blue”.
Expert Systems and Machine Learning
Expert systems are specialized for very specific and narrowly limited areas of application. One example is programs that convert computer tomographic images on the computer screen into three-dimensional images. Doctors can thus literally get a “picture” of the respective body part and its condition.
Self-learning systems are an important component of artificial intelligence. These are used, for example, in automated spam filters in the e-mail inbox. One feeds a computer with sample data, which it evaluates and analyzes.
The system recognizes patterns and similarities and can thus sort out spam mails, even if the sender or content is unknown to it. The human being only intervenes in a controlling manner and corrects, for example, if a mail is wrongly marked as spam. The computer remembers this in turn. The longer the system carries out these tasks, the better it becomes – a classic example of “machine learning”.
Voice recognition systems are also capable of learning. The more often they are used, the more they adapt to the linguistic peculiarities of the user. Over time, they can understand the user’s voice better and make fewer mistakes during processing.
Automatically into space
In 1997, machines in the service of mankind traveled to the planet Mars. The aim of the “Pathfinder Mission” was to bring scientific measuring equipment to the surface of Mars. Suitable techniques for flight phase, atmosphere entry, descent and landing should be developed and tested.
Everything had to work as automatically as possible, since human intervention from Earth is hardly possible due to the distance. A radio signal to Earth would take 14 minutes even if it were travelling at the speed of light.
But the “Pathfinder mission” was successful and thus laid the foundation for further Mars missions. In August 2012, the “Curiosity” vehicle landed on Mars: weighing 900 kilograms and equipped with a variety of instruments to explore the extent to which the planet is or was suitable as a biosphere.
Even the landing was spectacular: After entering the atmosphere, the probe automatically decelerated 20 meters above the surface and lowered “Curiosity” on ropes.
On Mars, “Curiosity” travels along with a plutonium drive, crushes and analyzes rocks with a laser and uses a gripper arm to pack rock samples into a microwave to melt them. Curiosity” has been on the move for more than eight years. It has already traveled more than twenty kilometers and is transmitting its findings to Earth. And his successor is also already on its way: “Perseverance” was sent into space in July 2020 (as of October 2020).