What is Artificial Intelligence: The Complete Explanation
Artificial intelligence, commonly abbreviated as AI, is the idea that a computer program or other machine can have intelligence equal to or greater than that of a human being. Although the basic concept of artificial intelligence is easy to grasp, the best definition of artificial intelligence is still hotly contested. For some, it’s enough that a machine can do something a human could do. Others, however, insist that the real criterion for an AI is that it can think rationally. Of course, there is also debate over whether a thinking machine is even possible or if even the most powerful artificial intelligence created in the future would ultimately just be a sophisticated piece of software.
AI as a concept was first conceived of by British mathematician Alan Turing (see the biography of Alan Turing) in a 1950 publication. The field was formalized at a series of meeting in Dartmouth, New Hampshire in 1956. In the decades since AI as a field has grown considerably and gone through many changes conceptually. Scientists studying AI have also struggled to scale their approaches up to be reproducible broadly speaking. Because of these issues, it is often difficult to pin down what exactly the purpose of AI should be or the best approach to get there. At the center of every definition, though, is an attempt to solve problems using machines that can use rational thought processes.
Today, artificial intelligence typically refers to specialized pieces of software used by companies, in healthcare, and in other settings that use pattern recognition and very large data sets to solve or assist with solving specific problems. For example, self-driving cars use road maps, GPS coordinates, and camera inputs to determine the best route from point A to point B and how to drive there safely. Present-day AIs may not meet every past definition of the concept but often excel at the specific issues they are designed for. Prior to their use, these AIs require significant “training” from human programmers and scientists and are far from perfect. In the future, technology may well advance to a point where machine intelligence can outperform humans and have thought processes completely indistinguishable from our own.
Artificial Intelligence: An Exact Definition
The best definition of artificial intelligence is hard to pin down. In the Stanford Encyclopedia of Philosophy, philosophers Bringsjord and Govindarajulu argue for a definition that approaches what AI is based on what it aims to achieve.
Less an exact definition and more of a framework, Bringsjord and Govindarajulu’s system split AI into a field that tries to create machines that meet one of the four following definitions:
- Machines that think like humans
- Machines that think rationally
- Machines that act like humans
- Machines that act rationally
These four aims can be further understood as a matrix, with each being either “human-based” or based in “ideal rationality” and either “reasoning-based” or “behavior-based.” So, for example, the first definition is a human-based and reasoning-based definition, while the fourth is ideal rationality and behavior-based.
Although it seems counter-intuitive to define a technological concept like artificial intelligence using a philosophy text, Bringsjord and Govindarajulu’s view of AI as a broad field that tries to achieve one of four specific aims allows for much more clarity than most definitions, which get bogged down in specifics and do not scale to general use. As a result, this framework is a great way to conceptualize the idea of “intelligent” artificial beings and arguably the best definition.
How does Artificial Intelligence work?
Artificial intelligence works by joining the high processing power of a specialized piece of computer software with incredibly large sets of data. Once an AI has been trained on the data set for a problem, it will be able to spot patterns in the data and make assumptions about relevant questions given to it.
In essence, what we call an “artificial intelligence” today is a program that can use super-human pattern recognition and high processing power to make decisions about specific problems.
How do you create Artificial Intelligence?
With today’s technology, the first step of creating an AI is to figure out what problem you want to solve. That’s because computers are not powerful enough yet to create a general AI that can solve any problem.
After you know your goal, the next steps are to gather data relevant to the problem and “train” the AI on it using either an existing algorithm for machine learning or one you have created on your own.
Once the AI has been trained on the data set, it is ready for implementation. In this final step, you build an environment for the AI to do its processing work. This can include selecting a specific programming language, installing hardware, or finding server space to host the finished software.
Where did Artificial Intelligence originate?
Alan Turing (see the biography of Alan Turing) was one of the first to clearly articulate it. In a 1950 paper titled Computing Machinery and Intelligence, Turing raises the question of whether machines can think and how their thinking might be tested. This paper also introduced what is now called the “Turing test,” a method to determine whether a machine can be considered intelligent.
Half a decade later, artificial intelligence as a modern field of study was conceived of at a conference held in Dartmouth, New Hampshire in 1956 called the Dartmouth Summer Research Project on Artificial Intelligence. At this conference, which lasted approximately 8 weeks, nearly fifty scientists including organizers Marvin Minsky, Nathaniel Rochester, and Claude Shannon debated the topic of how to create machines that could think and act for themselves.
Of course, it would be many decades before a machine could be created that might be considered an AI by any definition. All the same, the broad-ranging discussions held at this lengthy conference were incredibly influential in establishing the core concepts and ideas behind the field of artificial intelligence.
What are the applications of Artificial Intelligence?
There are numerous ways AI can be used to improve life for humans, especially for repetitive tasks or those that rely heavily on pattern recognition. Here are just a few of the most cited:
- Providing recommendations on e-commerce and streaming media websites
- Automatic translation systems
- Facial recognition software
- Analysis of satellite imagery to support research
- Assist technicians in healthcare settings by automating common tasks
Essentially, the applications of artificial intelligence are limited only by human imagination.
Examples of Artificial Intelligence in the real world
Regardless of whether you think AI is already among us or something that will only exist in the future, many companies and organizations around the world have embraced AI-like technologies.
AI imaging in healthcare
In a healthcare setting, artificial intelligence-driven software is routinely used to assist doctors and technicians when looking at radiograph scans for abnormalities such as tumors or cancerous growth. This use is described in a 2020 journal article in the Lancet, where authors Oren, Gersh, and Bhatt describe some ways the software could be improved by providing it with an understanding of the goal in scanning the images, as well as instructions on how to perform the scans themselves.
One of the most common examples of real-world AI is self-driving cars. These vehicles work by using cameras to detect obstacles and GPS to determine what route to take to get to a destination. By combining artificially intelligent software and specialized hardware, the cars can drive safely without much human input. In fact, some studies suggest that self-driving cars are even safer than regular cars driven by humans.
One of the most controversial uses of AI is its application in policing, where sophisticated programs are used to examine photographs of people’s faces and associate them with other photos. The artificial intelligence aspect of the software uses powerful pattern recognition to quickly match faces using the shape of and distance between facial features in a way that unaided humans would be unable to do.