Decoding Digital: Data Science, Machine Learning, and Artificial Intelligence
The world of digital--or business--transformation is full of terms that could be totally mystifying to a layperson. Ironically, many of them seem to imply that computers know better than we do. So machines can learn now? Intelligence is artificial? Data are doing science? It can all be very overwhelming. Let's dive into three terms all about knowing--so that we can all know a little more.
Artificial Intelligence - Execute
There are two main categories of artificial intelligence (AI), and you may be surprised to hear that the one most widely used is the least commonly talked about in popular culture. Narrow AI programs are those that are designed to "make decisions" or perform actions based on a predefined set of rules. They are designed to accomplish specific tasks, and those tasks do not include gaining consciousness and taking over the world. While they may be able to take in and respond to environmental input, as in the case of a smart speaker, the action they take based on that input follows the rules of the program. The program itself does not change in response to the environment. The more sensational type, General AI, gets a lot of attention in science fiction, but we are far from the point where a computer that is as smart as a human or smarter in every category could possibly become science fact.
Machine Learning - Predict
Machine learning programs have immense use as a tool for prediction. People are notoriously bad at guessing things accurately--the psychological heuristics we use to function in an overwhelming sea of sense data essentially make us into little piles of biases and preconceptions. Machine learning describes programs that are designed to model some outcome in a less wishy-washy way. Machine learning does have some biases--after all, a person had to create the initial rules and set the goal for the program, but the assumptions involved in a machine learning prediction are easier to know, since they based on coded rules and a set of input data, like AI. (In fact, machine learning is a kind of AI.) The difference with machine learning is that the coded rules are continuously trained on multiple sets of input data. The data is fed in, the model runs, and the results are judged to be successful or not. The program then takes the feedback and self-alters its rules such that they would turn out a more accurate result. This training process can be repeated hundreds of thousands of times, and even in the background as it performs the task it is designed to do, ever improving along the way, and as such the field of machine learning relies heavily on large pools of data.
Data Science - Investigate
It may be a bit misleading to include data science in the same list as machine learning and artificial intelligence. The first two terms referred to the apparent "intelligence" of computers, whereas data science is all about human knowledge and understanding. Data science is the field of analysis whereby a person analyzes data from the world around them in order to gain insights that can lead to better decisions in almost any area. Data scientists use tools to aggregate, clean, and organize data into a format that is most understandable to people. Then they can read the patterns they see in that data and draw intelligent interpretations from it. It takes a great deal of training and skill to usefully interpret the vast amounts of data that are available to us nowadays, but it's important to remember that while a data scientist may be a computer genius, driving the strategic success of your business is not a job for a computer. Thank your local data scientist today.