By Housni Antar | Staff Writer
It is quite evident that what is known as Artificial Intelligence is today’s current illusion. The reason I chose the word “illusion” is to signify the misrepresentation of AI (Artificial Intelligence). Through the ages that humankind passed through, we were given the chance to perceive multiple occurrences that one might call a “shift.” This shift represents a transition that dwells on increasing our capacity to experience reality from that single point known as perspective. Today the shift is AI and so is the illusion, for every shift is perceived in multiple ways, regardless of judging their correctness. “AI is going to change the world”, that statement is the result of the illusion, if one can rely on such statements, then it is preferable to alter it to another such as: “The algorithm of the latest artificial neuron is a pathway of unknown results, one is known…AI’s response to your questions.”. Hence, the purpose of this article is to introduce the reader briefly to the essence of today’s shift.
One can simply describe the AN (Artificial Neuron) as a learning algorithm that relies on its environment. An environment can mean multiple things including the input you provide when you use ChatGPT, the scanning of your thumb, or even the ability to identify a person from the camera of a grocery store. The artificial neuron is a mathematically based structure that has 3 main parts. The first part is the input that is given through a certain data channel. For the AN to satisfy its purpose, there must be multiple inputs that are organized based on “weights” (Suzuki, 2011). A weight is a value assigned to each input, or variable, to label the relevance of that input. The more relevant the input is the higher the weight (Suzuki, 2011). Now we have an n number of variables each with its unique w (weight). All weights are then multiplied by their corresponding variable. After the “multiplication phase,” the second phase would be the “summation” phase where everything is summed. We arrive at the final, and most important, step which is the “activation” which relies on an activation function (transfer function) (Suzuki, 2011).
A transfer function is used to collect the input that was reached based on the weights, and then through a mathematical process, derivation after summation, the result of the function related to the output of the AN will be achieved. The given answer of derivation will be relatively close to a single weighted input which becomes the output of the whole AN (Suzuki, 2011).
There are multiple functions used as a transfer function. The most used is the Sigmoid function (Suzuki, 2011). These functions are used to use the sum that was obtained and then give the result to be compared to some binary standard already placed. That binary standard is like the final decision, either a yes or a no (0 or 1). What determines which answer is chosen is a “threshold” that is relative to the chosen transfer function. For example, the derivative of the sigmoid function is bounded between –1 and 1, like a yes or a no. Of course, there would be a further pathway, a more complex mathematical functionality, but this might bring the image closer. After this process, we will obtain the output (Suzuki, 2011).
As stated earlier, this is but a mere introduction to ANNs (Artificial Neural Networks) and ANs. That is why today’s shift must be studied from all the aspects that affect human life, regardless of its negativity or positivity. AI has been rumored to be the reason humans will stop working. AI has also been rumored to drive humanity towards extinction. That is all broken down to the single ANNs that will be described below. The reason I chose to speak about the Recurrent ANN, as will be seen, and not the other types of ANN, is because it is the one that allows us to imagine the effect of “Learning.” By “Learning” I mean the ability to recognize what is observed as a data set that can be used to achieve other objectives. This is a straightforward description of “Machine Learning.” These are all topics that will affect the near future, and we have already started to witness their effect.
This a simple AN, Fig.1, but the real effect of this masterpiece of Raphael or Da Vinci, does not lie in its singularity, but in its polarity. The effect of multiple ANs connected to each other is the whole basis of AI. It is a matter of a massive number of comparisons to reach the final output that ChatGPT gives you, or whatever other form of AI. Having multiple ANs connected is called an Artificial Neural Network. There are multiple types of ANN (Artificial Neural Networks), but for the sake of simplicity, I will describe the one that is commonly relatable (Suzuki, 2011). The type is called a Recurrent ANN. When an ANN is established, its sole purpose is to provide a single output that is the result of multiple other outputs. A single AN provides a single output based on multiple inputs. Now, take that to the larger scale with multiple ANs connected, a simple ANN is called a Feed-Forward ANN. The Feed-Forward ANN describes a pathway between these ANs. A single AN’s output is sent to another AN as an input, and so this process continues until it reaches the final AN. The final AN receives the final inputs to give a single last output. The Recurrent ANN has the same concept applied with a twist. All ANs give their output to each other regardless of the pathway. That way the Recurrent ANN functions not just forward, but backward too (Suzuki, 2011). The concept can be understood through Fig.2.
Figures:
Fig.1
Source: Artificial Neural Networks – Methodological Advances and Biomedical Applications – Part 1 – Chapter 1- p. 3
Fig.2
Source: Artificial Neural Networks – Methodological Advances and Biomedical Applications – Part 1 – Chapter 1- p. 8
To return to the illusion, what we call the present is the mere presence of reality as an intangible gate to the future. What we call the past can be represented in simple words as a “data set.” That is all it takes for us humans to function, and that is also what an intelligent machine needs to function. This is unlike all the illusions that we have witnessed, this is an “intelligent” illusion. Will it change the future? Most people tend to answer such a question with multiple assumptions, others assume it is an actual illusion, if not a delusion. Perhaps the question that must be sought after is not related to the future but to the present. Thucydides once said “History repeats itself,” and what is our history, but a cycle of repetitive actions furnished with evolution. This leaves us with the question “What is the evolution that we sought after?”
References:
Suzuki, K. (2011, April). Artificial Neural Networks – methodological advances and biomedical … Research Gate. https://www.researchgate.net/publication/319316102_Artificial_Neural_Networks_-_Methodological_Advances_and_Biomedical_Applications