Artificial intelligence

Artificial intelligence (AI or A.I.) is a computer program or a machine that is able to learn and mimic human cognition.^[1]^[2] Sometimes, AI is used to talk about neural networks or deep learning.

Artificial intelligence is a system's ability to understand external data, to learn from that data, and to use what it has learned to achieve specific goals or tasks through adaptation.^[3]

Artificial intelligence is used a lot in language technology, speech recognition, image recognition, user interaction, and steering physical processes.

Origin of name

John McCarthy came up with the name, "artificial intelligence", in 1955. Intelligence allows an organism to act in a meaningful way in its environment. It includes the ability to get sensory inputs, and to react to these.

Laws

The European Union made a law (2024's second quarter) about artificial intelligence. It is the world's first law that regulates AI.^[4]

Connecticut and Colorado tried to pass laws (in 2024) regarding use of AI.^[5]

History

AI research started with a conference at Dartmouth College in 1956. It was a month-long brainstorming session many people who like AI came to. At the conference, they wrote programs which were able to beat humans at checkers or solving word problems. The Department of Defense started giving a lot of money to AI research, and labs were created all over the world.

In a paper on AI, mathematician James Lighthill wrote "no aspect of the discipline has so far seen discoveries generated the huge influence that was previously anticipated." The governments of the US and UK decided to spend money on other projects which caused little new research to be done. This was known as an "AI winter."^[6] In the 90s and early 2000s, AI became important again in data mining and medical diagnosis. This was possible because of faster computers and focusing on solving more specific problems. In 1997, the chess computer Deep Blue became the first computer program to beat chess world champion Garry Kasparov. In 2011, IBM Watson beat the top two Jeopardy! players Brad Rutter and Ken Jennings. In 2016, Google's AlphaGo beat top Go player Lee Sedol 4 out of 5 times.

The idea is perhaps much older. Julien Offray de La Mettrie (1709-1751) was a materialist thinker of the Enlightenment. In his work of 1748, L'Homme Machine, he had the idea that both matter and life organized themselves.^[7] He is seen as one of the precursors of Darwin's theory of evolution.^[8] Today, one field of artificial intelligence, called strong artificial intelligence wants to build a machine that can think like a person.^[9] However, weak artificial intelligence is about building a system that can support a human. One of the key problems is to make systems that can model uncertainity. Most of the time, this is done with probability theory and statistics.

Uses of artificial intelligence

Artificial intelligence is used in many different areas today.

Healthcare (see Artificial intelligence in healthcare) AI helps doctors diagnose diseases and find treatments for patients.
Industry-specific tasks; AI applications are used to solve problems in a workplace, industry, or institutions.
- Finance: Banks use AI to find fraud and make trading decisions.
Customer service: Many companies use AI chatbots to answer customer questions and provide support.
Personal assistants: AI assistants like Siri and Alexa help people manage their daily tasks.
Entertainment: AI is used in video games to create smart, responsive characters and stories that change in fun ways based on what the player does. Those subjects are part of Game artificial intelligence.

Military (Military artificial intelligence)
Generative AI (or Generative artificial intelligence)

Some software has become well-known, such as ChatGPT (a chatbot and virtual assistant).

Domains of artificial intelligence

There are different domains of artificial intelligence.

Pattern recognition like speech, writing, and handwriting
Knowledge engineering like logic programming and inference engines
Expert systems for question answering and Chatbots
Machine learning
Artificial neural networks and deep learning
Computer vision
Robotics
General game playing
Artificial life

Types (and classes) of artificial intelligence

Researchers Kaplan and Haenlein say there are three types of AI system: analytical, human-inspired, and humanized artificial intelligence.^[3]

Analytical A.I. has similarities with cognitive intelligence which tries to understand the world and make decisions based on that.
Human-inspired A.I. which tries to be more "human" with cognitive intelligence with emotional intelligence.
Humanized A.I. is able to understand human social activity and is able to be self aware^[10]

Other information

Researchers didn't know how difficult several issues were. They still couldn't offer computers things like emotions or common sense.

Faster computers, deep learning, and more data have made AI popular throughout the world.^[11]

An great intelligent machine is flexible and perceives what is around it. It would use what it learns to make its chance of success at some goal better.^[12]

AI has been successful with decoding human speech,^[1] playing games (like chess and Go), self-driving cars, and understanding complex data.^[13]

Someday, AI researchers hope to create computer programs that can learn, solve problems, and think logically.^[14]^[15] So far, most AI programs only do what computers can do well like searching databases or doing calculations. AI is not able to sense and understand what is happening around itself because of problems with what computers can do.

AI involves many different fields like computer science, mathematics, linguistics, psychology, neuroscience, and philosophy. Researchers hope to make a "general artificial intelligence" which can solve many problems instead of focusing on just one. Researchers are also trying to make creative and emotional AI which could create art. Alan Turing wrote in 1950 "I propose to consider the question 'can machines think'?"^[16] He said the question should be changed from if a machine "thinks" to "whether or not it is possible for machinery to show intelligent behavior".^[16] Alain Turing also created the Turing test. This is a very general test. If a human cannot tell if at the other end of the line, there is a machine or a human answering questions, the machine is intelligent.

The authors of Artificial Intelligence: A Modern Approach agree with Turing that AI must be defined by "acting" and not "thinking".^[17] But they don't think the test compares machines to people. "Aeronautical engineering texts do not define the goal of their field as making 'machines that fly so exactly like pigeons that they can fool other pigeons.Template:' "^[18] AI founder John McCarthy agreed, writing that "Artificial intelligence is not, by definition, simulation of human intelligence".^[19]

Computers can do some things like learning and problem solving, but not in the same way as people do.^[1]

Artificial intelligence programs create questions or gather questions from other places. If the program does not gather or create answers for questions, then the program is not artificial intelligence.

Some people also consider AI a danger to humanity.^[20]

AI and machine learning technology is used in applications including: search engines, recommendation systems, virtual assistants, autonomous vehicles, automatic language translation, facial recognition, image labeling, advertising, and driving internet traffic.

One interesting use of AI is in helping people with their personal relationships. For example, ChatGPT is a powerful AI tool that can create human-like text and assist in many communication tasks.

Optical character recognition is no longer thought of as an example of "artificial intelligence" since it's now commonly used.

Artificial Intelligence Media

An ontology represents knowledge as a set of concepts within a domain and the relationships between those concepts.
Kismet, a robot head which was made in the 1990s; a machine that can recognize and simulate emotions^[21]
Illustration of gradient descent for 3 different starting points; two parameters (represented by the plan coordinates) are adjusted in order to minimize the loss function (the height)
A simple Bayesian network, with the associated conditional probability tables
Expectation–maximization clustering of Old Faithful eruption data starts from a random guess but then successfully converges on an accurate clustering of the two physically distinct modes of eruption.
A neural network is an interconnected group of nodes, akin to the vast network of neurons in the human brain.
Machine learning as a subset of Artificial Intelligence
Vincent van Gogh in watercolour created by generative AI software
The first global AI Safety Summit was held in 2023 with a declaration calling for international co-operation.
The Turing test can provide some evidence of intelligence, but it penalizes non-human intelligent behavior.