Link Search Menu Expand Document
Publications
  1. What is Artificial Intelligence
    1. Deep Learning
    2. Neural Networks

What is Artificial Intelligence

When a person sees the term “artificial intelligence,” there may be many things that come to that individual’s mind. For the purposes of this research, it is important to define what artificial intelligence is and how some of its constituents work. Without the proper understanding of what “artificial intelligence” is, the ethics behind it cannot be fully understood. Artificial intelligence (known as “AI”) can be summed into the following definition: “artificial intelligence involves computational technologies that are inspired by – but typically operate differently from – the way people and other biological organisms sense, learn, reason, and take action” [1]. With the definition above, there are many things that can be molded around it.

The following lists some of the popular areas of artificial intelligence: ➢ Face recognition ➢ Search recommendations ➢ Chatbots ➢ Social media suggestions ➢ Advertisements ➢ Self-driving vehicles ➢ Image recognition ➢ Social network moderation

The above examples are already utilized in many areas of an individual’s life today, but how do they work? Deep learning is the driving force behind these ideas.

Deep Learning

Deep learning is a more specific area of artificial intelligence, but it is utilized in some way by most AI. Deep learning is defined as, “…representation-learning methods with multiple levels of representation, obtained by composing simple but non-linear modules that each transform the representation at one level (starting with the raw input) into a representation at a higher, slightly more abstract level” [2].

A way to help visualize the definition above is to look at image recognition, one of the largest areas of AI. Developing image recognition technologies includes showing a computer system a set of images and manually classifying those images. As the image is shown to the computer system, the algorithms will extract certain data from the image. The image can then be classified as a house, cat, dog, or of whatever the image consists. As more images are input into the computer system, the algorithm becomes more accurate and can better classify the images. As each image is shown, the machine outputs a vector of scores for each category. The scores help define the category to which the image most likely belongs. A function is developed to help calculate how much the vector scores were off from the desired vector scores. The machine will fine-tune different parameters to help get closer to the desired score. As more images are processed, the better the machine becomes at classification [2].

Neural Networks

Deep learning makes use of neural networks. Figure 1 helps show where neural networks sit within the scope of artificial intelligence. Neural networks are designed to mimic the way a human’s brain works, and they consist of four components: inputs, weights, bias, and output [3]. IBM provides the following formula and similar examples to help explain neural networks.

Figure 1- Breakdown of fields within artificial intelligence

Imagine an individual is considering whether to buy a gym membership. There are three factors to consider:

  1. Will it help with a healthier lifestyle? Yes[1]/No[0]
  2. Will it save money? Yes[1]/No[0]
  3. Will it help the individual lose weight? Yes[1]/No[0]

𝑥1= 1, exercising leads to a healthy lifestyle
𝑥2 = 0, gym memberships are expensive, and it would be cheaper to work out at home
𝑥3 = 1, exercising can reduce an individual’s weight

Weights must be assigned to the three options above to help clarify which ones are most important.

𝑤1 = 5; individual’s desire to be in shape
𝑤2 = 3, individual has a good job but is saving for a new car
𝑤3 = 2, individual wants to lose weight but not necessarily the most important

The bias provides an adjustable value by which the equation can be changed to best fit the data model [4]. If the input above is plugged into the equation, the output is one. This indicates that the individual should buy a gym membership. In the same manner that the above data can be calculated, neural networks plug-in image data to classify the image. The above is meant to serve as a basic understanding of how artificial intelligence works. Although neural networks are not used within every type of AI, they are still widely used. Different areas of where AI is applied are outlined in the next section. The above definitions and examples are used to help illustrate how AI might possibly work within different applications.

Table of Contents