Sidebar to the article Using Artificial Intelligence to Address Criminal Justice Needs, by Christopher Rigano. published in NIJ Journal issue no. 280.
1950: Alan Turing publishes his paper on creating thinking machines.
1956: John McCarthy presents his definition of artificial intelligence.
1956-1974: Reason searches or means-to-end algorithms were first developed to “walk” simple decision paths and make decisions. Such approaches provided the ability to solve complex mathematical expressions and process strings of words. The word processing is known as natural language processing. These approaches led to the ability to formulate logic and rules to interpret and formulate sentences and also marked the beginning of game theory, which was realized in basic computer games.
1980-1987: Complex systems were developed using logic rules and reasoning algorithms that mimic human experts. This began the rise of expert systems, such as decision support tools that learned the “rules” of a specific knowledge domain like those that a physician would follow when performing a medical diagnosis. Such systems were capable of complex reasoning but, unlike humans, they could not learn new rules to evolve and expand their decision-making.
1993-2009: Biologically inspired software known as “neural networks” came on the scene. These networks mimic the way living things learn how to identify complex patterns and, in doing so, can complete complex tasks. Character recognition for license plate readers was one of the first applications.
2010-present: Deep learning and big data are now in the limelight. Affordable graphical processing units from the gaming industry have enabled neural networks to be trained using big data. Layering these networks mimics how humans learn to recognize and categorize simple patterns into complex patterns. This software is being applied in automated facial and object detection and recognition as well as medical image diagnostics, financial patterns, and governance regulations. Projects such as Life Long Learning Machines, from the Defense Advanced Research Projects Agency, seek to further advance AI algorithms toward learning continuously in ways similar to those of humans.
About This Article
This article was published as part of NIJ Journal issue number 280, published January 2019, as a sidebar to the article Using Artificial Intelligence to Address Criminal Justice Needs, by Christopher Rigano.
[note 1] Alan Turing, “Computing Machinery and Intelligence,” Mind 49 (1950): 433-460.
[note 2] The Society for the Study of Artificial Intelligence and Simulation of Behaviour, “What is Artificial Intelligence.”
[note 3] Herbert A. Simon, The Sciences of the Artificial (Cambridge, MA: MIT Press, 1981).
[note 4] Daniel Crevier, AI: The Tumultuous Search for Artificial Intelligence (New York: Basic Books, 1993), ISBN 0-465-02997-3.
[note 5] Ibid.
[note 6] Pamela McCorduck, Machines Who Think, 2nd ed. (Natick, MA: A.K. Peters, Ltd., 2004), ISBN 1-56881-205-1, Online Computer Library Center, Inc.
[note 7] Navdeep Singh Gill, “Artificial Neural Networks, Neural Networks Applications and Algorithms,” Xenonstack, July 21, 2017; Andrew L. Beam, “Deep Learning 101 - Part 1: History and Background” and “Deep Learning 101 - Part 2: Multilayer Perceptrons,” Machine Learning and Medicine, February 23, 2017; and Andrej Karpathy, “CS231n: Convolutional Neural Networks for Visual Recognition,” Stanford University Computer Science Class.
[note 8] Beam, “Deep Learning 101 - Part 1” and “Deep Learning 101 - Part 2.”
[note 9] Karpathy, “CS231n.”
[note 10] Defense Advanced Research Projects Agency, “Toward Machines that Improve with Experience,” March 16, 2017.