No doubt each of us has come in contact with Artificial Intelligence (AI) — whether it is by shopping online and seeing ‘suggested for you’ products, or ads popping up in your Facebook feed, or at the bank when we are making a check deposit at the ATM. Other industries, ranging from health to fitness to media to dating apps to finance have all adopted AI in some capacity to optimize and automate their processes. Although known in the academic world since the 1990s, AI is only coming to mainstream utilization in recent years. So what exactly is the power of AI and why is it becoming so popular now?
AI is what is known as a forward model in computer science, meaning it is a computer model that makes decisions based on the input into the model, such as data that can be in the form of pictures, numbers, and really anything that is mathematically quantifiable. Thus, this type of model is able to modify its prediction based on the dynamic flow of input. So, the longer that the model receives input, the ‘smarter’ it becomes and the better guesses it can make about future behavior. It is different from the backward model, or non-AI predictive computer algorithms, in that in a backward model a prediction is made on a past data set and the unique parameters that are chosen at the time of the model building. Therefore, it does not have the capacity to modify its guesses or predictions because no matter how many times the backward model is fed data, it will always give the same answer for the same data point, but in a forward model, the answer changes with more data that is fed into the model. The reason why AI is being prevalently used now than before is twofold: 1) the science and the advancement behind algorithm development 2) with the abundance of data generated, it is the perfect time to use this model. One of the key principles of algorithm design for AI is that there must be enough ‘training’ data on which to ‘train’ the model before it is able to make meaningful predictions. This number can range from 10,000 data points and above. The abundance of data did not exist in the same capacity as it does now, and with the prevalence of high-speed computing, this is a perfect time utilize this model.