The term "Artificial Intelligence" was first mentioned at a scientific conference in the US city of Dartmouth in 1954. The scientist Marvin Minsky, who is considered one of the founding fathers of AI, defined the term in 1966 as follows: Artificial intelligence exists when machines do things that humans are assumed to have intelligence to do. At the end of the 1960s, the General Problem Solver was introduced. This was an AI system capable of solving simple problems. Also at the end of the 1960s, the ELIZA program developed at MIT attracted attention. The chat program was able to simulate a therapy conversation.
Better processor performance and memory capabilities ensured that AI capabilities continued to improve in the years that followed. In 2011, IBM introduced the Watson computer program. Watson was able to win against two human opponents on the quiz show Jeopardy. Users of computers or mobile devices now come into contact with artificial intelligence through programs such as Siri or Cortana. Siri and Cortana are intelligent assistants used in the iOS and Windows 10 operating systems, respectively. Methods for generating artificial intelligence
The basic assumption of AI is that human intelligence is the result of various calculations. In this context, AI itself can be generated in various ways. Meanwhile, there are AI systems whose main task is to recognize patterns and, as a result, perform appropriate actions. There are also the so-called knowledge-based AI systems. These attempt to solve problems using knowledge stored in a database. Other systems, in turn, employ methods from probability theory to react appropriately to given patterns.
Modern forms of AI
Among the most current expressions of artificial intelligence are approaches such as cognitive computing, neural networks and natural language processing. Cognitive computing is a concept that aims to adapt existing information systems to the requirements of today. In this way, the interaction between computer systems and humans is to be improved.
A neural network consists of artificial neurons and is based on the human brain in terms of its structure and functioning. As a result, a neural network is supposed to be able to produce particularly realistic calculations. Neural networks are now used in numerous areas of science and industry. For example, the company Google uses a neural network for its AI system DeepMind and combines it with methods from the field of machine learning. The goal of DeepMind and the machine learning approach is not only to equip computers with intelligence, but also to better understand how the brain works.
For this purpose, the so-called Deep Learning is also applied. Deep learning is a subarea of machine learning. Together with neural networks, Deep Learning currently offers the best way to recognize images and language. Natural Language Processing refers to the processing of natural language. Natural Language Processing, like Cognitive Computing, focuses on the interaction between computer and user.
Possible applications and limitations
There are now a wide range of possible applications for AI systems. Companies often take advantage of the opportunity to make their communication with customers more efficient by using chatbots. Warehouse management or purchases can now also be handled by AI-based systems. Artificial intelligence in the form of robotics is used in the production of machines or devices. In addition, AI can also be used in the automotive sector. There, for example, artificial intelligence is used to develop and implement self-driving cars. Although AI is now useful in many areas, its use is always associated with problems and risks.
A Google research group pointed out the dangers associated with the use of AI in March 2016. To this end, the researchers formulated a catalog of questions aimed at specifying possible security risks of self-learning and intelligent systems. Among other things, the researchers asked how a robot can explore its environment in such a way that it does not endanger people.