May 1997, New York, USA. The Deep Blue chess computer developed by IBM is the first computer to defeat the reigning chess world champion Garri Kasparov under tournament conditions with 3.5:2.5.
February 2011, US television show “Jeopardy”. Watson”, developed by IBM and several universities, beats Jeopardy masters Ken Jennings and Brad Rutter. It uses a four terabyte knowledge base that includes the full text of Wikipedia.
March 2016, Seoul, South Korea. The computer program AlphaGo competes against South Korean Lee Sedol, currently considered the strongest Go player, and wins the tournament with 4-1, playing moves that reveal completely new options of the game and provide the Go scene with unexpectedly new knowledge.
These three events, taken from a multitude of achievements, outline the history of AI, artificial intelligence. The next series of blog entries aims to investigate artificial intelligence at a relatively high, i.e. rough, level and to answer the following questions, among others:
- What is artificial intelligence?
- What are their current applications?
- Which technologies are used?
- Who develops artificial intelligence, which companies or universities are largely behind it?
- What influence does artificial intelligence have on our future?
The information comes from various textbooks, articles and publications, which are cited as sources. Since the topic of artificial intelligence can no longer be broken down into a few blog entries, this series should only serve as a rough overview of this subarea of current research.
In this first entry we look at the definition of artificial intelligence. The name itself raises one of mankind’s biggest questions: “What is intelligence?”. How is it possible that simple matter can build a complicated system of connections with which it can receive, understand, predict and even manipulate information from an even more complex world?
While this question is far from being answered, but rather represents a major aspect of today’s research, the concept of “artificial intelligence” goes even further. “Artificial” implies something non-natural, that is, something created by man. In order to produce something like this, however, we first need the necessary knowledge. This gives us an outlook on the AI research project:
The task is to understand intelligence and to build and apply it oneself.
Since such a project involves all cognitive activities, a precise definition of artificial intelligence is rather difficult. Instead, we want to collect the descriptions of artificial intelligence and get a rough idea of what it actually means.
The term “artificial intelligence” was first introduced in 1955 by one of the pioneers of the AI, John McCarthy:
“AI’s goal is to develop machines that behave as if they had intelligence.”
In the Encyclopedia Britannica, however, one finds the following definition:
“AI is the ability of digital computers or computer-controlled robots to solve tasks that are normally associated with intelligent beings…”
However, in view of the complexity of intelligence, both definitions appear rather weak. There are machines that seem to show intelligent behavior, but are far from true intelligence. One would certainly expect some intelligent being behind complicated calculations, but nevertheless, a computer with some computing power does not represent an artificial intelligence.
A concise definition of the goal of artificial intelligence is provided by Elaine Rich in 1983:
Artificial Intelligence is the study of how to make computers do things at which, at the moment, people are better.
This definition can also be found in the textbook “Artificial Intelligence: A Modern Approach” by Stuart Russell and Peter Norvig. There, eight different definitions of artificial intelligence are divided into four different categories:
In the field of human action, for example, the Turing Test (Alan Turing, 1950) can also be found. A machine is questioned for 5 minutes by a person in another room and passes the test if, in at least 30% of the cases, the person does not notice that he is communicating with a machine. In the total Turing test a video transmission takes place additionally, necessary abilities include among other things: Understanding and interpreting language, applying stored knowledge, automated reasoning, machine learning, computer vision and robotics
Cognitive science plays an important role in the field of human thought. Before one can imitate or recreate human thinking, one must understand how the brain works. The topic of artificial intelligence in this field is therefore strongly linked to the experimental investigation of humans and animals.
The field of rational thinking is based exclusively on logical conclusions and classifications. This approach has the difficulty of translating any knowledge into formal logic and keeping up with the necessary computing power.
Finally, the area of rational action also takes into account the limits of logic. It is taken into account that there are situations in which no strict logical conclusion can be drawn and yet action is required. Rationally acting artificial intelligence therefore also considers other rational methods as pure reasoning.
One last way to define artificial intelligence is to consider the capabilities of the program. Three main areas are distinguished here:
- Artificial Narrow Intelligence: A program focused on a specific activity.
- Artificial General Intelligence: A program or machine that reaches a human level in a wide range of activities.
- Artificial Superintelligence: This type of AI was defined more precisely by philosopher Nick Bostrom: “an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills.”
At the moment the programs we have created are still in the first of the three stages. In the next blog entries we will see, however, that we are getting closer and closer to the border to strong artificial intelligence with programs such as AlphaZero.
In summary, the term “artificial intelligence” thus reflects the attempt to imitate and surpass the peculiarity of man, namely his natural intelligence, on the one hand through logical and rational calculations, but on the other hand also through a systematic approach to human consciousness itself.
The technologies used for this purpose are used in a wide variety of areas. In the next blog entry we will examine some of them in more detail.
Sources and further information:
Book series and textbook on artificial intelligence (main sources for this blog entry):
- Computational Intelligence, W. Bibel, R. Kruse, B. Nebel, https://www.springer.com/series/12572
- Russell, Stuart, and Peter Norvig. Artificial Intelligence: A Modern Approach, Global Edition, Pearson Education Limited, 2016.
Articles, papers and blog entries:
- https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html (in the second part there is a larger collection of helpful links)