How can AI have the ability to be asked a question and respond in a completely legitimate way in a matter of seconds? When thinking about neural networks, they work very similarly to the human brain. Imagine when you ask AI a question that there is a series of different lightbulbs that are each in rows and columns. Say there are two lights in the first row and one blinks for yes, one blinks for no, and both blink for unsure. This is what is referred to as an input node. Now imagine the next row contains a lightbulb combination for each different possible answer for any given question. This is the idea for neural networks, featuring an ever-growing database of nodes to be able to answer any given question. These can create large combinations of possibilities, triggering different “lightbulbs” that each lead to a different final output. For reference, if each light had the state of on or off in a 5 × 5 grid, there would be a total of 33,554,432 different possible combinations. Neural networks are modeled as simplified brains that in some ways can be considered even stronger. They have different weights attached to nodes to attempt to match the strength of synapses in the brain. Additionally, they both have the ability to learn, like neural networks through backpropagation compared to the brain’s ability to form connections of neurons. The one thing that neural networks truly miss is the ability to utilize common sense; they rely entirely on their neural node pathways and can make simple mistakes that a human could spot from a mile away. The takeaway here is that most of neural networks is formatted through probability, leading to whichever answer has the highest percentage of being correct.