The term artificial intelligence, or AI can be described as an Anthropomorphism (Anthropomorphism is the attribution of human traits, emotions, and intentions to non-human entities and is considered to be an innate tendency of human psychology). Artificial intelligence, neural networks, deep learning, these phrases are an example of anthropomorphism.
A summary of what was observed at U of T Neural Networks and Machine Learning classes: Teaching the machine to learn means it will interpret data in a method set by programming. One way to do this is take a block of information and break it down into the smallest components. Then we read the small components information and sometimes do a comparison using an equation to tell us the difference between the information.
In computer language, the smallest data is "true" or "false", or '1' or '0'. In the processor, the brain of the computer, 1 means 'on' and '0' means 'off'. This refers to the physical component, the circuitry of the processor.
One example of this is facial recognition software, akin to Facebook's software. Take a close up photograph (black and white) of two faces. Zoom into the pixel level, and we compare the color value of pixel at location (1,0) using a Cartesian plane of Headshot A with the color value of the pixel at the same location in Headshot B. The result is two numbers to compare, and this is the data. A rule defined by a programmer could be that if the color value (range is black to white, or 0-255) was less than 50% of the color value for white, that this color is considered dark, and color values less than 25% of 255 are black and nearly black. This gives us 4 possible values.
The next step, interpreting the data, requires some math skills. A programmer writes a type of calculation called an algorithm to determine how the computer will do this. The computer reads the color of all the pixels, and decide what percent difference in values between a pixel or a group of pixels could mean. Using the shade values as percentiles, a pixel with a value of 64 or less would be black/nearly black, and from 64-128 would be nearly dark, 128-191 would be nearly white, and 191-255 white. There’s 4 shades of color. Now we get the computer to determine where the shapes start and end. We ‘teach’ it to identify a series of pixels with the same shade all sharing a border. This is a mathematical or logical formula, determining if the amount of ‘nearly white’ or ‘white’ surrounding the group is enough to be considered ‘yes, it’s a white area’. The programmer sets the threshold for a positive or negative determination by giving a number, and so has written the formula.
Now we have the faces mapped out as information that the computer can apply math to and so make a comparison. we use a similar formula to compare Headshot A and Headshot B. You can think of it as having the headshots printed with black ink on two plastic sheets, and layering them one on top of the other and examining them. From here we can see if the black ink is in the same spots on each plastic sheet. Pretty easy comparison to do with the human eye. Our brains also determine what the likelihood these two pictures match from the information.
Any kind of math can be applied to any kind of data; the math is what we call AI. We’re feeding a huge number of mathematical variables into a calculator, and getting the numbers that come out the other end, which we’ve already defined the meaning. AI is computing, which means it’s another way of doing complicated math in an increasingly efficient way.