Artificial intelligence

from Wikipedia, the free encyclopedia

Artificial intelligence ( KI ), also artificial intelligence ( AI or A.I. ), English artificial intelligence ( AI or AI ) is a branch of computer science that deals with the automation of intelligent behavior and machine learning . The term cannot be clearly defined as there is already a lack of a precise definition of “ intelligence ”. Nevertheless, it is used in research and development.

General

In general, artificial intelligence refers to the attempt to reproduce certain human decision-making structures, e.g. B. a computer is built and programmed in such a way that it can deal with problems relatively independently. Often, however, it is also used to denote an imitated intelligence, whereby "intelligent behavior " is to be simulated by mostly simple algorithms , for example with computer opponents in computer games .

In the understanding of the concept of artificial intelligence often resulting from the reflected Enlightenment originating idea of "man as machine" resist whose imitation of the so-called strong AI is the goal: to create an intelligence that the human mind is to mechanize, or a machine to design and build that reacts intelligently or just behaves like a human . After decades of research, the goals of strong AI are still visionary.

Strong and weak AI

Strong AI would be computer systems that can take on the work of completing difficult tasks at eye level with people . In contrast, weak AI is about mastering specific application problems. Human thinking and technical applications are to be supported here in individual areas. The ability to learn is a main requirement of AI systems and must be an integral part that cannot be added afterwards . A second main criterion is the ability of an AI system to deal with uncertainty and probabilistic information. In particular, those applications are of interest which, according to general understanding, seem to require a form of "intelligence" to solve them. Ultimately, the weak AI is about the simulation of intelligent behavior using mathematics and computer science, it is not about creating consciousness or a deeper understanding of intelligence. While the creation of strong AI has failed to this day due to its philosophical question , significant progress has been made on the side of weak AI in recent years.

A strong AI system doesn't have to have a lot in common with humans. It will probably have a different cognitive architecture and its developmental stages will also not be comparable to the evolutionary cognitive stages of human thought ( evolution of thought ). Above all, it cannot be assumed that an artificial intelligence possesses feelings such as love, hate, fear or joy. However, she can simulate behavior corresponding to such feelings.

Research areas

In addition to the research results of core computer science itself, research on AI has included results from psychology , neurology and neurosciences , mathematics and logic , communication science , philosophy and linguistics . Conversely, research on AI also influenced other areas, especially neurosciences. This can be seen in the training in the field of neuroinformatics , which is assigned to biology-oriented computer science, as well as computational neuroscience .

In artificial neural networks , there are techniques that have been developed from the mid-20th century and on the neurophysiology build.

AI is therefore not a closed research area. Rather, techniques from different disciplines are used without these having to be linked to one another.

An important meeting is the International Joint Conference on Artificial Intelligence (IJCAI), which has been taking place since 1969.

history

Sub-areas

Knowledge based systems

Knowledge-based systems model a form of rational intelligence for so-called expert systems . These are able to provide answers to a user's question on the basis of formalized specialist knowledge and the logical conclusions drawn from it . Exemplary applications can be found in the diagnosis of diseases or the search for and elimination of errors in technical systems.

Examples of knowledge-based systems are Cyc and Watson .

Pattern analysis and pattern recognition

Visual intelligence makes it possible to recognize and analyze images or shapes . Examples of application include handwriting recognition , identification of people through facial recognition , comparison of fingerprints or the iris , industrial quality control and production automation (the latter in combination with findings from robotics).

Using linguistic intelligence , it is possible, for example, to convert a written text into speech ( speech synthesis ) and, conversely, to write down a spoken text ( speech recognition ). This automatic language processing can be expanded so that meaning can be assigned to words and texts using latent semantic analysis ( LSI for short ).

Examples of systems for pattern recognition are Google Brain and Microsoft Adam .

Pattern prediction

Pattern prediction is an extension of pattern recognition. It represents the basis of the hierarchical temporal memory defined by Jeff Hawkins .

“Prediction is not just one of the things your brain does. It is the primary function of the neocortex, and the foundation of intelligence. "

Prediction isn't just one of the things your brain does. It is the main function of the neocortex and the foundation of intelligence. "

- Jeff Hawkins : On Intelligence

Such systems offer the advantage that, for. B. not only is a certain object recognized in a single image (pattern recognition), but also a series of images can be used to predict where the object will be next.

robotics

The robotics deals with manipulative intelligence. With the help of robots , dangerous activities such as mine hunting or the same manipulations, such as B. can occur during welding or painting can be automated.

The basic idea is to create systems that can understand the intelligent behavior of living things. Examples of such robots are ASIMO and Atlas .

Modeling based on the entropy force

Based on the work of the physicist Alexander Wissner-Gross , an intelligent system can be modeled using the entropy force . Here, an intelligent agent tries its surroundings (state X 0 ) by an action (force field F to influence) to the greatest possible freedom of action (entropy S ) in a future state X to achieve.

Artificial life

AI overlaps with the discipline artificial life ( Artificial Life , AL) is seen as a parent or as a sub-discipline. AL must integrate their knowledge, since cognition is a core property of natural life, not just of humans.

Methods

The methods of AI can be roughly classified into two dimensions: symbolic vs. neural AI and simulation method vs. phenomenological method. The following graphic illustrates the relationships:

Artificial Intelligence Methods.png

Neural AI takes a bottom-up approach and aims to simulate the human brain as precisely as possible. Conversely, symbolic AI follows a top-down approach and approaches intelligence performance from a conceptual level. The simulation method is based as closely as possible on the actual cognitive processes of humans. In contrast, the phenomenological approach only depends on the result.

Many older methods that were developed in AI are based on heuristic solution processes. In recent times, mathematically sound approaches from statistics , mathematical programming and approximation theory have played an important role.

The specific techniques of AI can be roughly divided into groups:

Search

AI often deals with problems that require specific solutions. Different search algorithms are used. A prime example of search is the process of finding the way , which plays a central role in many computer games and is based on search algorithms such as the A * algorithm .

To plan

In addition to searching for solutions, planning is an important aspect of AI. The planning process is divided into two phases:

  1. The formulation of goals : A goal is defined based on the current state of the environment or world. A goal here is a set of world states in which a certain goal predicate is fulfilled.
  2. The formulation of the problem : After it is known which goals are to be striven for, the formulation of the problem determines which actions and world states are to be considered. There are different types of problems here .

Planning systems plan and create sequences of actions from such problem descriptions, which agent systems can carry out in order to achieve their goals.

Optimization methods

AI tasks often lead to optimization problems. Depending on the structure, these are solved either with search algorithms from computer science or, increasingly, with means of mathematical programming . Well-known heuristic search methods from the context of AI are evolutionary algorithms .

Logical closing

One of the questions posed by AI is the creation of knowledge representations that can then be used for automatic logical reasoning . Human knowledge is - as far as possible - formalized in order to bring it into a machine-readable form. The developers of various ontologies have dedicated themselves to this goal .

Early on, AI was concerned with constructing automatic proof systems that would help mathematicians and computer scientists to prove sentences and to program ( logic programming ). Two difficulties emerged:

  1. If you formulate sentences in the natural language, relatively comfortable descriptive languages, the resulting search problems become all too time-consuming. In practice, compromises had to be made in which the description language was a little more complicated for the user, but the associated optimization problems were easier to handle for the computer ( Prolog , expert systems ).
  2. Even powerful descriptive languages ​​become unwieldy when trying to formulate uncertain or incomplete knowledge. For practical problems, this can be a serious limitation. Current research therefore examines systems that apply the rules of probability theory to explicitly model ignorance and uncertainty. In terms of algorithms, these methods differ from the older methods: in addition to symbols , probability distributions are also manipulated.

Another form of logical inference, the induction is ( inductive inference , inductive logic ) are generalized in the examples of rules ( machine learning ). Here, too, the type and power of the knowledge representation play an important role. A distinction is made between symbolic systems in which the knowledge - both the examples and the induced rules - is explicitly represented, and sub-symbolic systems such as neural networks, which are "trained" to behave in a predictable manner, but which do not allow any insight into the solutions that have been learned .

Approximation methods

In many applications, it is a matter of deriving a general rule from a set of data ( machine learning ). Mathematically, this leads to an approximation problem . In the context of AI, artificial neural networks were proposed, among other things, which can be used as universal function approximators, but are difficult to analyze, especially with many hidden layers. Sometimes alternative methods are used that are mathematically easier to analyze.

Artificial neural network

Artificial intelligence has recently made great strides in the field of artificial neural networks, also known as deep learning. Neural networks that are roughly inspired by the structure of the brain are artificially simulated on the computer. Many of the recent successes such as handwriting recognition , speech recognition, face recognition , autonomous driving , machine translation , including the success of AlphaGo, are based on this technology.

Applications

There are numerous areas of application for artificial intelligence. Some examples briefly summarized:

AI in medicine

AI in law

A large part of the work of lawyers consists of analyzing files , for example precedents, in order to develop arguments from them. This kind of work can now be partly done by AIs. Consulting firm McKinsey estimated in 2017 that around 22 percent of the work of lawyers and 35 percent of the work of legal assistants could be automated with the help of AIs. The AIs are trained on millions of documents and case studies and legal applications. Then an AI can mark the documents that a lawyer needs for his case; often better than a human could. JPMorgan announced that it will use AI Contract Intelligence, which, according to JPMorgan, can analyze a lot of data in seconds that would take lawyers and legal assistants 360,000 hours.

AI in marketing

Artificial intelligence is used in marketing, for example, to send advertising emails, to replace customer service with social bots and chatbots , to carry out analyzes and forecasts of the market and the customer, for example on the basis of big data , and to create customer-specific advertisements, recommendations and search results, as well as developing programmed processes. In March 2018, the online mail order company Zalando intended to cut 250 jobs in the marketing area in Berlin, which are to be replaced by artificial intelligence.

AI in computer and board games

In computer games, an AI is mostly used to control NPC, so-called non-player characters , who simulate human-like behavior (for example as simulated allies or computer opponents ) or to control certain things in the game world or in the functions of the game character (for example route finding , procedural generation , automatic improvements and completions in route construction or other algorithms). In some games, the level of difficulty of the AI ​​opponents can be set and optionally you can choose whether you want to play against an AI, against real players or a mixed form. With a few games, the AI ​​can automatically adapt to the game behavior or can learn from mistakes. Since there are often no opponents in single player mode , an AI is used. In addition, AI is used in computer games to simulate many or very special characters that are difficult or impossible to adopt by real people. In some cases, however, AIs can also be simply tricked into computer games because a person can circumvent a certain pattern of an AI. The realism and gameplay of a computer game is therefore often measured against the AI.

AI is also used in strategy board games as a substitute for the human partner. Even world champions have little chance of winning against very powerful versions of these programs. AI achieved success against human professional players, for example in backgammon , chess , checkers and go . Mastering complex games is often the subject of research in order to develop and demonstrate new methods of artificial intelligence. These programs are now playing games among themselves. At the end of 2017, the new development AlphaZero clearly triumphed against the world's best chess program Stockfish in 100 games. In addition, AIs are also being developed that control video games such as jump 'n' runs , role play or racing games instead of a human player . Similarly, the development of e-sports range, try in the professional gamers to beat the best AIs, while developers work to defeat the best players by AI.

AI to create works of art

Researchers from Tübingen have used neural networks to paint a given photo in the style of a famous artist. B. Van Gogh or Edvard Munch. Researchers at Google have trained neural networks to use a type of white noise to produce images in the style of Van Gogh and other artists. The pictures were later auctioned off.

In July 2017, researchers at Rutgers University presented an AI that produces artistic paintings. The AI ​​was trained with around 80,000 works of art from Western art history. In a blind test, the paintings created by the KI were mixed with pictures that had been exhibited at the Art Basel art fair and 18 test subjects (artistic laypeople) were submitted to a blind test for assessment. The test subjects were asked to assess whether the images were generated by humans or a computer. In the case of real works of art exhibited at Art Basel, the test subjects assumed that 52% of all works were created by a computer. For the AI-based images, the test subjects assumed this for only 25% of all images.

In March 2018 a video work of art was published in which an Ornella Muti created by AI acted. With the help of an artificial neural network, the artist Joseph Ayerle calculated new film sequences that the real Italian actress never played.

In October 2018, the auction house Christie's auctioned the " Portrait of Edmond de Belamy " created by artificial intelligence . The painting, originally estimated to have a market value of $ 7,000 to $ 10,000, generated auction proceeds of $ 432,500.

Behind the creation of the portrait was the French artist group Obvious, who had trained an artificial intelligence with the image data of 15,000 real paintings from the 14th to the 20th century. Special attention was paid to the fact that the picture was not signed with the artist's signature, but with the formula "min G max D Ex [log (D (x))] + Ez [log (1-D (G (z )))] "which, according to the artist team, was used when it was created.

Author George RR Martin was writing his sixth book in the Game of Thrones series , which was eagerly awaited by the fan base. The programmer Zack Thoutt trained an AI (Recurrent Neural Net) with the first five books in the series and had the AI ​​write a sixth book. The result was published on the Internet in summer 2017. The AI ​​developed individual characters just as it was expected in some fan theories without the AI ​​knowing about it. There are flaws in the grammar, individual characters who have already died reappear and the storylines are not very exciting.

Sunspring is the first short film (2016) whose script was written by an AI.

In its Magenta project, Google tries to create AIs that are creative. In the summer of 2017, a piano improvisation was presented that was composed by an AI. In the summer of 2016, the Magenta project released a short pop song that was composed by an AI.

The music of the album "I am AI" by the singer Taryn Southern , presented in autumn 2017, was composed by an AI. To compose a song with the help of an AI, you use software such as Amper Music or Jukedeck, select the genre and other parameters such as length of the song, instrumentation, etc. The AI ​​then composes a unique song within seconds. A musician can then put pieces of these examples together to create a song of their own. So everyone can create more or less professional music. More and more musicians admit to using AI as a tool when composing. Skygge's album “Hello World” was also composed entirely with an AI (flow machine). The AI ​​composes sound pieces, which are then sorted, selected and put together by humans, the so-called curating.

The view of the artists involved in the discourse on the role of AI as the originator of a work of art is controversial. The motto of the artist group Obvious is: "Creativity is not just for people." On the contrary, there is the statement of the artist Joseph Ayerle, who is quoted by the Massachusetts Institute of Technology as saying: "AI can create, but it is not creative" .

AI to produce product design

A team from the US 3D software expert Autodesk and the well-known designer Philippe Starck have jointly created - according to the information provided by those involved - the first "chair developed jointly by artificial intelligence and humans", the so-called AI Chair .

Turing test

In order to have a criterion when a machine simulates an intelligence equivalent to humans , Alan Turing suggested the Turing test named after him. A person asks any questions to another person or an AI via a terminal without knowing who is answering. The questioner then has to decide whether the interviewee is a machine or a human. If the machine is indistinguishable from humans, then, according to Turing, the machine is intelligent. So far, no machine has been able to pass the Turing test without a doubt. The Loebner Prize for the Turing Test has existed since 1991 .

Technological singularity

It is roughly understood as the point in time at which artificial intelligence outperforms human intelligence. From this point on, further development is mainly driven by AI and no longer by humans.

Super intelligence

A superintelligence describes a being or a machine with an intelligence that is superior to humans in many or all areas. The term is often used in the field of artificial intelligence.

Comparing Artificial Intelligence with Human Intelligence

According to Wolfgang Wahlster , human intelligence has to be divided into different areas: cognitive intelligence, sensorimotor intelligence, emotional intelligence , and social intelligence .

Cognitive intelligence

When it comes to cognitive intelligence, cognitive systems are already superior to humans in many areas. This area includes the game of chess, the game of Go and other board games. Ultimately, it is about absorbing and learning about knowledge as well as combining and drawing conclusions from this knowledge. This often corresponds to what people acquire in an academic training.

Sensorimotor intelligence

With this intelligence, humans are still superior to machines, but some machines are superior in areas of individual sensors. Basically, the human eye is very well trained. But a suitable video camera can also process light in the infrared and UV range, which a human cannot. In acoustics, microphones can pick up significantly lower volumes or in frequency ranges than the human ear. This applies even more to the sense of smell and taste, where machine sensors are clearly superior. However, a person can combine these sensory impressions (sensor fusion), which a machine has so far only been able to do little. However, this could change within a few years.

Emotional intelligence

So far, the machine has done almost nothing in this area. A person can empathize with another person, feel sympathy and empathy, compassion, pity, sadness, fear, joy, write love poems, have outbursts of anger, etc. What machines can, however, already rudimentary, is the so-called sentiment analysis , i. H. by observing human body language, ie the face, gestures, etc., “reading” a person's emotions.

Social intelligence

This is the ability to (re-) react appropriately in a human group, for example to recognize a mood or to influence it constructively, e.g. B. the team spirit. A skill that is mostly developed by entrepreneurs but also by politicians. So far, the machine has not been able to do anything in this area.

Artificial Intelligence Awareness

It is a basic assumption in neuroscience that consciousness is a product of our brain (see Neural Correlate of Consciousness ). According to Jürgen Schmidhuber , awareness is only a by-product of problem-solving in the brain. Artificial problem solvers (e.g. autonomous mobile robots ) are also advantageous if they are “aware” of themselves and their surroundings. Schmidhuber's “consciousness” in the context of autonomous robots refers to a digital world model including the system itself, but not to the experience of states. A world model could be learned in the context of reinforcement learning by rewarding actions that expand the world model.

Adjacent Sciences

Linguistics

The interpretation of human language by machines plays a crucial role in AI research. Any results of the Turing test result primarily in dialogue situations that have to be mastered.

With its grammar models and psycholinguistic semantic models such as feature or prototype semantics, linguistics provides the basis for the machine “understanding” of complex natural language phrases. The central question is how language signs can actually have a meaning for an artificial intelligence. The Chinese Room argument of the philosopher John Searle should show that it would be possible to pass the Turing test even if the language characters used are not given any importance. In particular, results from the embodiment area also emphasize the relevance of experiences that are based on the embodiment of an agent and its integration into a meaningful environment for every form of cognition, i.e. also for the construction of meaning by an intelligence.

An interface between linguistics and computer science is the computational linguistics , which deals among other things with machine language processing and artificial intelligence.

psychology

Psychology deals, among other things, with the term intelligence .

psychotherapy

In psychotherapy research, experimental applications of artificial intelligence have existed for some time to bridge deficits and bottlenecks in psychotherapeutic care and to save costs, but also to identify impending crises in patients on the waiting list at an early stage.

philosophy

The philosophical aspects of the AI ​​problem are among the most far-reaching in all of computer science.

The answers that are given to the central questions in this area extend far into ontological and epistemological topics that have preoccupied human thinking since the beginning of philosophy. Anyone who gives such answers must also draw the conclusions from them for people and themselves. It is not uncommon to want to proceed in reverse and apply the answers that were found before the development of artificial intelligence to them. But as it turned out, artificial intelligence has prompted numerous researchers to look at problems such as the relationship between matter and spirit , the origins of consciousness, the limits of knowledge, the problem of emergence , the possibility of extra-human intelligence, etc. in a new light and to reevaluate in part.

A perspective that is committed to metaphysical or idealistic thinking considers it impossible (in the sense of a weak AI) that machines could ever have more than just simulated consciousness with real knowledge and freedom. From an ontological point of view, the American philosopher Hubert Dreyfus criticizes the concept of strong AI. Building on the ontology of the “worldliness of the world” developed by Martin Heidegger in his work Being and Time , Dreyfus tries to show that one cannot go back to the phenomenon of the world as a meaningful totality of meaning: meaning, i. H. Relationships between things in the world are an emergence phenomenon, because there is no such thing as “some sense” and then “more sense”. With this, however, the task of programming meaningful relationships between things in the world into a computer also turns out to be an actually impossible or infinite undertaking. This is because meaning cannot be created by adding initially meaningless elements.

An evolutionary-progressive school of thought, on the other hand, sees it as possible (in the sense of a strong AI) that systems of artificial intelligence could one day surpass humans in what is currently considered specifically human. On the one hand, this harbors the risk that such AI machines could turn against people's interests. On the other hand, this technology offers the opportunity to solve problems that humans find difficult to solve due to their limited capacities (see also technological singularity ).

Further points of contact can be found in analytical philosophy .

In addition to the question of being and consciousness, the question of whether an AI can be held responsible for its illegal actions or misconduct (e.g. in a car accident caused by an autonomous vehicle) arises within the framework of legal philosophy and robot ethics who is responsible for everything. Developers are confronted with the question of how an AI acts morally and ethically correctly. For example, people are considering how to solve the trolley problem with autonomous vehicles.

In his three robot laws, the Russian-American biochemist and non-fiction author Isaac Asimov describes the prerequisites for peaceful and supportive coexistence between AI and humans. These laws were later extended by other authors.

Human rights

The central questions in the use of AI include the division of legal obligations between states and companies as well as the implications of human rights with regard to the use of AI in certain areas of application, e.g. B. in face recognition or facilitating the decision-making of courts. The extent of technological cooperation in the field of AI with states that do not adhere to basic human rights standards will also be discussed from an economic ethical and international law perspective.

Computer science

The Artificial Intelligence is meshed closely with the other disciplines of computer science. A differentiation can be attempted on the basis of the results obtained. To this end, it makes sense to differentiate between different dimensions of intelligence:

  1. The ability to process any symbol (not just numbers).
  2. The construction of an inner model of the outer world, a self-model, as well as the relationship between self and world.
  3. The ability to apply knowledge appropriately.
  4. The ability to uncover the relationships contained in the stored knowledge, d. H. to be able to reason logically.
  5. The ability to generalize (abstraction) and specialize (i.e. to apply general relationships to concrete facts).
  6. The ability to transfer acquired knowledge and experience to new, previously unknown situations.
  7. The ability to behave according to plan and to be able to form appropriate strategies to achieve the goals.
  8. Adaptability to different, u. U. Situations and problematic environments that change over time.
  9. Ability to learn combined with the ability to assess partial progress or regression.
  10. The ability to act in blurred or incompletely described or recognized situations.
  11. The ability to recognize patterns (possess sensors) and actively deal with the environment (possess effectors).
  12. Have a means of communication as complex and expressive as human language.

Criticism of AI research

Stephen Hawking warned against AI in 2014 and saw it as a threat to humanity. The end of mankind could be ushered in through AI. The future will show whether the machines will take control at some point. But it is already clear today that machines are increasingly displacing people from the labor market.

In August 2017, 116 entrepreneurs and experts from the technology industry (including Mustafa Suleyman , Elon Musk , Yoshua Bengio , Stuart Russell , Jürgen Schmidhuber ) demanded in an open letter to the UN that autonomous weapons should be banned or the CCW, which has existed since 1983 List should be set. The Certain Conventional Weapons are banned by the UN and include chemical weapons. After gunpowder and the atomic bomb, the third revolution of warfare threatens. Quoting from the letter: "Once this Pandora's box is opened it becomes difficult to close again" and "Once invented, it could allow armed conflict on an unprecedented scale and faster than humans can grasp". Terrorists and despots could use the autonomous weapons and even hack them.

Such positions were argued against a. Rodney Brooks and Jean-Gabriel Ganascia .

In February 2018 a report by a project group of leading experts in the field of AI was published, which warns of possible "malicious uses of artificial intelligence" (English original title: "The Malicious Use of Artificial Intelligence"). Researchers from the universities of Oxford, Yale and Stanford, as well as developers from Microsoft and Google were involved. The report makes reference to existing technologies and uses various scenarios to demonstrate how these could be abused by terrorists, criminals and despotic governments. The authors of the report therefore call for closer cooperation between researchers, developers and legislators in the field of AI and propose concrete measures on how the dangers of abuse could be reduced.

The philosopher Richard David Precht opposes the idea that in the future there is a threat of bad will or the striving for power on the part of a developed artificial intelligence; rather, the potential danger lies in their incorrect use.

Suggestions for dealing with AI

Microsoft President Brad Smith suggested that a code of conduct , such as a Digital Geneva Convention , be drawn up to reduce artificial intelligence risks.

In the context of the use and programming of artificial intelligence, the ethicist Peter Dabrock recommends not only increasing the digital competence of those involved, but also using classic educational elements. Knowledge of religion, literature, mathematics, foreign languages, music and sport are a good prerequisite for coping with the associated challenges and increasing the ability to differentiate and recognize ambiguities.

Dissemination of AI in Germany

The number of companies that use AI technologies is still relatively low in Germany. At the end of 2018, only 6 percent of companies were using or implementing AI. 17 percent stated that they test or at least plan to use AI. The ZEW study also comes to a similar result. In 2019, around 17,500 companies in the reporting group of the innovation survey (manufacturing industry and predominantly business-oriented services) used AI in products, services or internal processes. That is 5.8 percent of the companies in the reporting group.

The AI ​​observatory

With the Observatory Artificial Intelligence in Work and Society (abbreviated to: KI Observatory), a project by the Digital Work Society think tank, the Federal Ministry of Labor and Social Affairs is focusing on the question of the effects of AI on work and society. The AI ​​observatory acts at the interface between politics, science, business and society: it acts as a knowledge carrier and initiator. The AI ​​observatory has the task of anticipating the effects of AI in the world of work at an early stage and identifying the need for action. In this way, the work unit, which started in March 2020, is making a contribution to the implementation of the goals formulated in the Federal Government's AI strategy - for example, the safe and common good use of AI. In addition, the AI ​​observatory is intended to empower and empower various social actors in dealing with artificial intelligence with the help of dialogue and participation formats.

The observatory's specific tasks are set out in the five fields of action:

1. Technology foresight and technology assessment

2. AI in labor and social administration

3. Framework for AI / social technology design

4. Development of international and European structures

5. Social dialogue and networking

Representation in film and literature

AI has been dealt with in art, film, and literature since Classical Modernism . The artistic processing - in contrast to AI research, in which the technical realization is in the foreground - is primarily about the moral, ethical and religious aspects and consequences of a non-human, “machine intelligence”.

During the Renaissance , the term homunculus was coined, an artificial miniature human without a soul. Human-like automatons appeared in literature in the 18th and 19th centuries, for example in ETA Hoffmann's Der Sandmann and Jean Paul's Der Maschinenmann .

In the 20th and 21st centuries, science fiction takes up the topic in many ways in film and prose . In 1920 the writer Karel Čapek coined the term "robot" in his play RUR ; In 1926, Fritz Lang discussed robots in Metropolis , which do human work.

In the various works, the film audience was presented with the robots as intelligent and differentiated machines with very different personalities: they are developed to use them for good purposes, but often turn into dangerous machines that develop hostile plans against people. In the course of film history, they are increasingly becoming self-confident beings who want to submit to humanity.

The Lithuanian artist republic of Užupis is another form of artistic engagement with AI . In its Munich embassy, ​​the artificially intelligent research humanoid “Roboy” acts as a consul and the constitution contains its own article on artificial intelligence (“Any artificial intelligence has the right to believe in a good will of humanity [The Munich Article]. ”).

Examples (selection)

Social impact

In the course of the industrial revolution , muscle power was replaced by the machine with the invention of the steam engine ( horsepower for watts ). The digital revolution could replace human thinking with machine AI.

The American entrepreneur Elon Musk predicts that in the future there will be less and less gainful employment that cannot be done better and cheaper by a machine, which is why fewer and fewer workers would be needed. As a result of the largely mechanical production, the products and services would become very cheap. In this context, he supports the introduction of an unconditional basic income. The physicist Stephen Hawking said: It is already clear today that machines are increasingly displacing people from the labor market. Microsoft founder Bill Gates sees the development similarly. He calls for a robot tax to be able to cope with the social tasks of the future.

The computer scientist Constanze Kurz stated in an interview that technical progress has always existed. In the past, however, technical change usually took place over generations, leaving enough time to train for new tasks. Today, technical change takes place within a few years, so that people do not have enough time to train themselves for new tasks. The spokesman for the Chaos Computer Club , Frank Rieger , warned in various publications (e.g. the book Arbeitsfrei ) that the accelerated automation of many work areas will lead to more and more people losing their jobs in the near future (e.g. truck drivers by self-driving cars ). Among other things, there is a risk of weakening trade unions, which could lose members. Rieger therefore advocates a “socialization of the automation dividend”, ie taxation of non-human work, so that general prosperity grows and is distributed fairly through the growth of the economy in the form of a basic income.

In a study in 2013, scientists at Oxford University tested a large number of jobs for their automation. The scientists divided the jobs into different risk groups. 47 percent of the jobs examined in the USA were classified in the highest risk group, i.e. This means that the risk of these jobs being automated within the next one or two decades (as of 2013) is very high.

Jack Ma , the founder of the Chinese Internet company Alibaba , warned in a lecture that people should prepare for significant changes in the labor market because AI will change the world. In the past 200 years, manufacturing and services have created the jobs. But now because of the AIs and the robots, hardly any jobs will be created there. Jack Ma criticized today's school education (he used to be an English teacher). The students are not being trained for the needs of tomorrow, but rather for an economy that will soon no longer exist. The schools would train the unemployed of tomorrow. It doesn't make sense to want to compete with the AIs and robots. Schools should train students to be as innovative and creative as possible. Jack Ma assumes that the AIs will destroy many jobs, but also create many new jobs. The question is whether students are being trained for these new jobs.

Jürgen Schmidhuber answered the question of whether AIs will soon overtake us or whether we will have to worry about our jobs: “Artificial intelligences will learn almost everything that humans can - and much more. Your neural networks are becoming smarter from experience and, due to the rapidly becoming cheaper hardware, a hundred times more powerful every ten years. Our formal theory of fun even allows curiosity and creativity to be implemented to build artificial scientists and artists. ”And“ Every five years, arithmetic becomes 10 times cheaper. If the trend continues, small computers will soon be able to calculate as much as a human brain, 50 years later like all 10 billion brains put together. ”Schmidhuber sees the necessity as a consequence of the inevitable progressive automation and the associated loss of gainful employment an unconditional basic income. “Robot owners will have to pay taxes in order to feed the members of our society who no longer have existentially necessary jobs. Anyone who does not support this to a certain extent invokes the human versus machine revolution. "

Erik Brynjolfsson is of the opinion that the emergence of radical parties in the USA and Europe is the result of the fact that many people can no longer keep up with technological progress. When people lose their jobs, those people get angry, Brynjolfsson said. He, too, thinks that in the future most jobs will be done by machines.

In a speech to Harvard graduates, Mark Zuckerberg said that the introduction of an unconditional basic income was necessary. Something could no longer be right if, as a Harvard dropout, he could make billions within a few years, while millions of university graduates could not pay off their debts. What is needed is a basis on which everyone can be innovative and creative.

In November 2017, Deutsche Bank boss John Cryan announced a major downsizing. The company employs 97,000 people. 4,000 jobs have already been cut in the last 12 months. In the near future, 9,000 more jobs are to be cut. Half of all jobs are to be cut in the medium term. Cryan justified this step by stating that the competition is already providing comparable performance with around half of its employees. Cryan said, "We do too much manual labor, which makes us error-prone and inefficient". The company could become much more efficient, especially through machine learning or artificial intelligence. Many bankers worked like robots anyway, Cryan said. Qualified machines should take the place of qualified employees, says Cryan.

In November 2017, the futurologist Lars Thomson predicted huge upheavals in technology, work, values ​​and society over the next 10 years. In 2025, a household robot could set the breakfast table, clean windows, take over care services, etc. which would destroy jobs. Today there are already 181 companies worldwide that are working on smart robots. The price of such a robot is around 20,000 euros today. The artificial intelligence market will be larger than the automotive market in a few years. How quickly 10 years passed, you would see if you look back 10 years when the first smartphone came onto the market. He regrets that hardly anyone in our society recognizes this development, which will completely change our society. In 10 years' time, robots will take over the work of today's maids in hotels. The advantage for the hotel manager: the robot does not want any wages, no days off, does not have to be taxed or insured. The disadvantage: the state no longer receives taxes and people are unemployed. Therefore, there will be no getting around an unconditional basic income and the introduction of a robot tax. Thomson sees the danger of a division in society if the pace of change exceeds people's ability to change. At the same time, the AI ​​will free people from work. Society must define guard rails for the AIs.

In an interview in January 2018, the CEO of Google Sundar Pichai said that the current development of artificial intelligence is more important for the development of mankind than the discovery of fire and the development of electricity. With the current development of AI, no stone will be left unturned. It is therefore important that society deals with the topic. This is the only way to limit the risks and exploit the potential. Google is currently one of the leading companies in the field of AI. The AI ​​assistant from Google alone is already installed on hundreds of millions of Android smartphones. But AI is already being used billions of times in search engines. DeepMind, a company acquired by Google, rushes from milestone to milestone and so on in AI research. a. with AlphaGo , AlphaGo Zero, AlphaZero .

The Institute for Employment Research (IAB), which is part of the Federal Employment Agency, demonstrated in a study from 4/2018 which human work can be replaced by machines in Germany. The study comes to the conclusion that in 2016, 25 percent of paid human activities could have been done by machines, which corresponds to around 8 million jobs in Germany. An earlier study came to a value of 15 percent for 2013. Most affected are manufacturing occupations with around 83 percent, but also company-related service occupations with 60 percent, occupations in company management and organization with 57 percent, occupations in agriculture, forestry and horticulture with 44 percent, etc. Compared to 2013 and 2016 are particularly strong Logistics and transport professions increased (from 36 to 56 percent), an area in which around 2.4 million people are employed in Germany. Overall, the study assumes that in the near future, 70 percent of human paid activities could be taken over by machines. Machines could e.g. B. take over: incoming goods inspection, assembly inspection, order picking, insurance applications, tax returns, etc. The techniques that drive these changes are: artificial intelligence, big data, 3-D printing and virtual reality. Even if there would be no layoffs, employees must at least expect major changes in their job description and thus major relearning. New professional fields will also emerge. Also, not everything that is already possible today will be implemented, and certainly not immediately. A factor in this delay are ethical and legal aspects, but also the high costs of automation. Artificial intelligence is not always cheaper than human intelligence.

In a guest post in February 2018, SAP CEO Bill McDermott said that people would fear the changes that a world with robots and AIs would bring. A first milestone was the victory of the Deep Blue machine over the reigning world chess champion Gary Kasparov in 1997. Another milestone was the victory of the Watson machine over humans in the quiz show Jeopardy in 2011. And the next big step was the victories of AlphaGo and its successors AlphaGo Zero and AlphaZero in 2016 and 2017. The profound changes that AI would also bring in the workplace are now on everyone's lips. In order to avoid any negative effects of the new technologies on society, it now required well thought-out planning. Public authorities, the private sector and education must work together to equip young people with the skills they need in the digital economy. Retraining and lifelong learning are the new normal today. Jobs would not be completely replaced by machines, but mostly in parts. Many new jobs would also be created. Economic development would be fueled by the AI. For 2030, value added in the range of 16 trillion dollars and a growth in gross domestic product of 26 percent are expected. Automation could save companies $ 3 to $ 4 trillion annually in the future.

On June 28, 2018, the German Bundestag set up a study commission on Artificial Intelligence - Social Responsibility and Economic Potential , which is to present a final report with recommendations for action by summer 2020.

Film documentaries

literature

Audio

Web links

German

English

Individual evidence

  1. ^ A b Nils J. Nilsson: The Quest for Artificial Intelligence. A History of Ideas and Achievements . Cambridge University Press, New York 2009.
  2. ^ Nick Bostrom: Superintelligence. Scenarios of a coming revolution . Suhrkamp, ​​2016, p. 42.
  3. ^ Nick Bostrom: Superintelligence. Scenarios of a coming revolution. Suhrkamp, ​​Frankfurt am Main. 2016, p. 50 f.
  4. ^ Daniela Hernandez: Microsoft Challenges Google's Artificial Brain With 'Project Adam'. In: Wired . July 14, 2014, accessed August 5, 2014 .
  5. Jeff Hawkins, Sandra Blakeslee: On Intelligence . Owl Books, 2005, ISBN 978-0-8050-7853-4 , pp. 89 .
  6. Alexander D. Wissner-Gross, CE Freer: Causal Entropic Forces. (PDF) In: Physical Review Letters. Institute for Applied Computational Science ( Harvard University ), The Media Laboratory ( MIT ), Department of Mathematics ( University of Hawaiʻi at Mānoa ), April 19, 2013, accessed on August 8, 2014 .
  7. Alex Wissner-Gross: A new equation for intelligence. In: YouTube . TED , February 6, 2014, accessed August 5, 2014 .
  8. ^ Mark A. Bedau: Artificial life: organization, adaptation and complexity from the bottom up . In: Department of Philosophy, ReedCollege, 3023 SE Woodstock Blvd., Portland OR 97202, USA (Ed.): Trends in Cognitive Sciences . tape 7 , no. 11 . Portland, OR, USA November 2003 ( reed.edu [PDF; accessed March 12, 2019]).
  9. Wolfgang Banzhaf, Barry McMullin: Artificial Life. In: Grzegorz Rozenberg, Thomas Bäck, Joost N. Kok (Eds.): Handbook of Natural Computing . Springer, 2012, ISBN 978-3-540-92909-3 .
  10. Artificial intelligence is revolutionizing astronomy. science.ORF.at, December 15, 2017, accessed on March 12, 2019 .
  11. Lukas Staffler, Oliver Jany: Artificial intelligence and criminal justice - an orientation . In: Journal for International Criminal Law Doctrine . tape 2020 , p. 164-177 ( zis-online.com [PDF]).
  12. Sascha Mattke: Artificial intelligence displaces human work in legal professions. Technology Review, December 20, 2017, accessed March 12, 2019 . heise.de from December 20, 2017.
  13. Bernd Mewes: AI-controlled marketing: Zalando cuts 250 jobs. In: Heise online . March 10, 2018 . Retrieved March 13, 2018.
  14. Reality and future: This is how AI is used in marketing . In: Internet World Business . ( internetworld.de [accessed on March 13, 2018]).
  15. Peter Gentsch: Artificial Intelligence for Sales, Marketing and Service - With AI and Bots to an Algorithmic Business - Concepts, Technologies and Best Practices . Springer, 2018, ISBN 978-3-658-19146-7 .
  16. ^ Klaus Breuer: Programming Computer Games: Artificial Intelligence for Artificial Brains . Oldenbourg Wissenschaftsverlag, Munich 2012, ISBN 978-3-486-71789-1 .
  17. Matthias Kreienbrink: Artificial Intelligence: The game knows what you're going to do . In: Zeit Online . December 11, 2017 ( zeit.de [accessed on March 16, 2018]).
  18. Artificial intelligence in games: AI is as intelligent as its developer . In: Golem.de . ( golem.de [accessed on March 16, 2018]).
  19. ^ Georgios N. Yannakakis, Julian Togelius: Artificial Intelligence and Games . 2018, doi : 10.1007 / 978-3-319-63519-4 ( springer.com [accessed December 22, 2018]).
  20. Jörg Breithut: Artificial Intelligence AlphaZero: In four hours to the world chess champion . In: Spiegel Online . December 8, 2017 ( spiegel.de [accessed March 16, 2018]).
  21. "Mario Kart" learns artificial intelligence from an older man. Retrieved March 16, 2018 .
  22. The Physics arXiv Blog: Neural Net Learns Breakout Then Thrashes Human Gamers. In: Medium. December 23, 2013, accessed March 16, 2018 .
  23. A video game-playing AI beat Q * bert in a way no one's ever seen before . In: The Verge . ( theverge.com [accessed March 16, 2018]).
  24. Eike Kühl: Artificial Intelligence: Now it also defeats professional gamers . In: Zeit Online . August 19, 2017 ( zeit.de [accessed March 16, 2018]).
  25. Dainius: New Neural Algorithm Can 'Paint' Photos In Style Of Any Artist From Van Gogh To Picasso. boredpanda.com, 2016, accessed March 12, 2019 .
  26. Alyssa Buffenstein: Google's Artificial Brain Creates Its Own Artworks and They Are Freaky. news.artnet.com/art-world/, June 22, 2015, accessed on March 12, 2019 (English).
  27. ^ Sarah Cascone: Google's 'Inceptionism' Art Sells Big at San Francisco Auction. news.artnet.com/, March 2, 2016, accessed on March 12, 2019 .
  28. Christian Gall: Can computers also produce art? Retrieved February 25, 2020 .
  29. ^ Claire Voon: Humans Prefer Computer-Generated Paintings to Those at Art Basel. hyperallergic.com/, July 31, 2017, accessed on March 12, 2019 .
  30. Katerina Cizek, William Uricchio, Sarah Wolozin: PART 6: MEDIA CO-CREATION WITH NON-HUMAN SYSTEMS . In: Collective Wisdom . PubPub, June 3, 2019 ( mit.edu [accessed February 14, 2020]).
  31. FAZ.net with DPA: Christie's sells AI art: min G max D Ex [log (D (x))] + Ez [log (1-D (G (z)))] has painted something . In: FAZ.NET . ISSN  0174-4909 ( faz.net [accessed February 12, 2020]).
  32. DER SPIEGEL: Christie's achieves 432,500 dollars with AI painting - DER SPIEGEL - Netzwelt. Retrieved February 12, 2020 .
  33. Alexander Armbruster: Computer writes the sixth book of Game of Thrones. faz.net, August 30, 2017, accessed March 12, 2019 .
  34. a b Schmidhuber: "Our robots show feelings" October 1, 2017.
  35. ^ Script written by KI: Sunsspring. A Sci-Fi Short Film Starring Thomas Middleditch. video Sunspring, June 9, 2016, accessed on March 12, 2019 .
  36. Tomislav Bezmalinovic: Google wants to teach computers to compose and make jokes. Mixed Reality News & Podcast, September 4, 2017, accessed March 12, 2019 .
  37. Vera Bauer: Google Magenta team publishes first piece of music composed with AI. mobilegeeks.de, June 4, 2016, accessed on March 12, 2019 .
  38. Everyone can compose with artificial intelligence deutschlandfunkkultur.de from December 21, 2017.
  39. KI will rock you zeit.de from December 26, 2017.
  40. Artificial intelligence can now also pop (well, almost) gruenderszene.de from February 8, 2018.
  41. Süddeutsche Zeitung: Art by Algorithm: Christie's auctions AI paintings. Retrieved February 14, 2020 .
  42. Katerina Cizek, William Uricchio, Sarah Wolozin: PART 6: MEDIA CO-CREATION WITH NON-HUMAN SYSTEMS . In: Collective Wisdom . PubPub, June 3, 2019 ( mit.edu [accessed February 14, 2020]).
  43. Katharina Cichosch, DER SPIEGEL: KI-Design: This chair was designed by a computer (with the help of design star Philippe Starck) - DER SPIEGEL - style. Retrieved February 17, 2020 .
  44. ^ Alan Turing: Computing Machinery and Intelligence. Out: Mind No. 236 October 1950.
  45. Artificial Intelligence: Overhyped or Underestimated? - CeBIT future talk, March 14, 2016 .
  46. Robots have to pay taxes Interview with Jürgen Schmidhuber In: wiwo.de , January 31, 2016.
  47. Stevan Harnad : The Symbol Grounding Problem . In: Physica D. 42, 1990, pp. 335-346.
  48. Franz-Josef Hücker: The Pygmalion mythology in psychotherapy. In: Psychotherapy Forum. Vol. 16, No. 3, 2008, (Springer), (Vienna), pp. 128-135.
  49. Nora Saskia Görg et al .: Predicting dropout in patients receiving Dialectical Behavior Therapy. Accessed August 27, 2018 (English).
  50. Cf. Hubert Dreyfus: In-der-Welt-sein und Weltlichkeit: Heidegger's Critique of Cartesianism. In: Thomas Rentsch: Being and Time . Akademie Verlag, Berlin 2001, p. 69ff.
  51. heise online: When computers make the difference between life and death: Who is liable if the AI ​​kills? Retrieved on March 20, 2018 (German).
  52. Tanja Oppelt: Ethics Commission presents results in Berlin: Self-driving cars and morality . Bayerischer Rundfunk, June 20, 2017; accessed on August 20, 2019.
  53. Christoph Stockburger: Autonomous driving: What should your car do now? Spiegel Online , August 29, 2016; accessed on March 20, 2018.
  54. Alexander Kriebitz and Christoph Lütge: Artificial Intelligence and Human Rights: A Business Ethical Assessment Business and Human Rights Journal , January 2020; accessed on July 28, 2020.
  55. David Kaye: Report of the Special Rapporteur to the General Assembly on AI and its impact on freedom of opinion and expression OHCHR , January 2020; accessed on July 28, 2020.
  56. a b Hilal Kalafat: Physicist warns of artificial intelligence. In: Handelsblatt . 3rd December 2014.
  57. a b Stephen Hawking warns of artificial intelligence ( memento from July 18, 2015 in the web archive archive.today )
  58. ^ Rory Cellan-Jones: Stephen Hawking - will AI kill or save humankind? In: BBC News . October 20, 2016 ( bbc.com [accessed October 28, 2018]).
  59. Elon Musk and 116 experts call for a ban on killer robots , t3n.de.
  60. Elon Musk and Co. warn of killer robots , faz.net.
  61. Gero von Randow : Artificial Intelligence: Too intelligent for life. In: The time . September 14, 2017. Retrieved September 27, 2017 .
  62. a b c Miles Brundage, Shahar Avin, Jack Clark, Helen Toner, Peter Eckersley, Ben Garfinkel, Allan Dafoe, Paul Scharre, Thomas Zeitzoff, Bobby Filar, Hyrum Anderson, Heather Roff, Gregory C. Allen, Jacob Steinhardt, Carrick Flynn , Seán Ó hÉigeartaigh, Simon Beard, Haydn Belfield, Sebastian Farquhar, Clare Lyle, Rebecca Crootof, Owain Evans, Michael Page, Joanna Bryson, Roman Yampolskiy, Dario Amodei: The Malicious Use of Artificial Intelligence. (PDF) Center of the Study for Existential Risk, February 20, 2018, accessed on March 9, 2018 .
  63. Robots have no morals. Why the talk of super-intelligent, all-powerful machines is just a big red herring. In: Die Zeit , June 18, 2020, p. 32.
  64. Peter Dabrock: We should rely on classical education. In: Awakening Artificial Intelligence - What it means and how it changes our lives. Google LLC, SZ Scala GmbH, 2018, p. 34.
  65. PwC: Artificial Intelligence in Companies . 2019 ( pwc.de [PDF]).
  66. Federal Ministry for Economic Affairs and Energy: Use of Artificial Intelligence in the German Economy. Retrieved June 22, 2020 .
  67. AI observatory. Retrieved June 22, 2020 .
  68. The 5 fields of action of the AI ​​observatory. Retrieved June 22, 2020 .
  69. Lisa Xanke, Elisabeth Bärenz: Artificial Intelligence in Literature and Film - Fiction or Reality? , Online article from the University of Karlsruhe, accessed on July 20, 2012, p. 1.
  70. Xanke, Bärenz, p. 37
  71. a b Xanke, Bärenz, p. 38.
  72. a b Xanke, Bärenz, p. 39.
  73. uzhupisembassy.eu accessed on October 28, 2018.
  74. ARD Quarks and Co: Out of Control - When Computers Take Power, 2016 Minute 16:30, September 6, 2016.
  75. video: Interview with Elon Musk: Elon Musk says Universal Basic Income is “going to be necessary.” 19th February 2017.
  76. Elon Musk: Unconditional Basic Income Is Inevitable February 19, 2017.
  77. ARD alpha: Constanze Kurz: Total automation, 2014
  78. Frank Rieger, Constanze Kurz : Arbeitsfrei: A journey of discovery to the machines that replace us .
  79. Frank Rieger: Robots have to secure our pension . In: FAZ , May 18, 2012.
  80. The Future Of Employment: How Susceptible Are Jobs To Computerization? (PDF) oxfordmartin.ox.ac.uk, September 17, 2013.
  81. Jack Ma says 'stop training kids for manufacturing jobs' September 21, 2017.
  82. Press agency APA / sda: Robot researcher advocates unconditional basic income . In: diepresse.com , January 15, 2017; Retrieved April 7, 2017.
  83. Jürgen Schmidhuber: We have to raise robots like children. Interview by Vinzenz Greiner, January 15, 2017.
  84. ARD: Quarks: Out of Control? When computers take over . Minute 16:50 and 19:30, ard.de, September 6, 2016; accessed on October 1, 2017.
  85. Mark Zuckerberg explains why everyone should receive an unconditional basic income . businessinsider.de, May 26, 2017; accessed on October 15, 2017.
  86. ^ Mark Zuckerberg's Commencement address at Harvard . news.harvard.edu, November 25, 2017; accessed on October 15, 2017.
  87. Machines instead of employees: Deutsche Bank boss holds out the prospect of considerable job cuts . faz.net, November 9, 2017; accessed on November 10, 2017.
  88. Researcher predicts dramatic change , In: schwaebische.de , November 9, 2017. Accessed November 10, 2017.
  89. Google boss: Artificial intelligence “more important than fire and electricity” , vrodo.de from January 20, 2018.
  90. Labor market effects of digitization by 2035 , iab.de from 4/2019
  91. These jobs are particularly threatened by robots , welt.de of February 16, 2018.
  92. SAP boss McDermott: AI will soon bring trillion sales , produktion.de from February 26, 2018.
  93. Lisa Brüssler: German Bundestag - Enquete Commission for Artificial Intelligence used . In: German Bundestag . ( bundestag.de [accessed September 6, 2018]).