What is the difference between Machine Learning and Artificial Intelligence?
Artificial intelligence has been around much longer then Machine Learning. The idea has been discussed since the computers invention, around the 1950’s. The general definition of AI is the science of making machines imitate the human thought process. In other words, the system’s ability to do things that would otherwise require human intelligence such as reasoning, perception, movement, processing language, knowledge, learning, achieving social skills, creativity, and planning.
In discussions on this topic, Mateo Restrepo the Head of Data Analytics of Yuxi Global explained that there are two types of AI; the stronger one, which although we haven’t reached this point yet, the desired result is to create a machine that replicates human intelligence. The second is the weaker AI, where the system or “agent” has the capability of making decisions as we do or at times can make better ones.
An example of this would be an automatic or an autonomous car. They are knowledgeable operators and due to their decision-making ability, they consider themselves “intelligent”. They are aware of when and where to stop, they recognize pedestrians, they guide themselves within the lanes and they work with multiple operating systems. These “intelligent agents” process many things as we humans do, they use sensors that gather information about their surrounding and outside world, and to act on that environment they must have “actuators” who control those actions, that is to say that, just like humans, these agents have a way to execute the decisions they make and thus be able to act on the real world.
Machine Learning is a specific subset of AI. It’s the science that involves the development of self-learning algorithms. It can include neural networks and it processes statistics to develop self-learning algorithms. Simply put, through developing a specific process these algorithms learn from datasets to complete tasks and solve problems more precise than AI.
On the other hand, a problem that would be for ML is to predict whether a customer of a bank will pay a loan or not (using the information it has compiled on equity and hundreds of different variables)
Let’s Imagine …
Let’s imagine that we need to organize the automated shifts of a group of 100 employees of a supermarket. There are certain parameters the schedule much adhere to: the time frame of the peak hours of the store, which is from 5:00 to 7:00, for example or making sure the employees who work the night shift get the required wages. For an individual it can be a nearly impossible task to balance so many variables (as there are many specific conditions that need to be met). This is called constraint satisfaction and it is a problem solved with AI and not in ML due to the fact that ML is based on formulating a large amount of possible solutions to a specific problem.
With ML a machine learns from statistical samples, it learns to develop the best functional relationship between x and y, it learns the form of the f and recognizes that x calculates y, but this learning process can only be done through examples, thousands of examples.
Machine Learning is fueled by data while Artificial Intelligence learns from the algorithms that engineers develop, and whose task it is to make them the most generic and manageable possible so that it can be applied to the greatest amount of problems with the least amount of readjustments. In short, make them adaptable.
The problem
The main misconception between ML and AI is that AI wants to build a model that performs well, but it has to figure out its subproblems; In the case of automatic cars, it has to recognize different factors such as which signal is being viewed, it then has to classify that signal and decipher its enviroment, and through this process its able to pinpoint exactly what needs to be resolved. It is through ML that each subproblem is solved. Picture AI as a universe that cannot function without ML solving individual categories. You use ML techniques to find solutions to problems that come up in AI.
In general
Artificial intelligence implies machines think as humans do. There is a level of complexity, however, to determining whether a machine is really “thinking” or not, so its assumed that: in order to create a good AI model, the program must simulate and be good at doing all the things we humans are good at.
It’s safe to say most industries and companies are investing in and incorporating AI in their ventures and products, and for someone unfamiliar with the topic it is might be impossible to realize the interaction we have with it through trivial activities we do on a day to day basis, like when we use our smartphone or surf the web.
One of the types of AI that we interact with the most is machine learning.
Companies with an online presence use Machine Learning to create recommendations on internet search engines. It’s no coincidence when certain e-commerce web sites, products or services show up more often on the users’ screen and it’s based on the data collected of individual behavior when we choose a movie or buy a garment. All these recommendations are built from user classification and analyzing click patterns.
If you need to resolve this kind of problems, Yuxi Global can help you. Contact us and see how your solutions can get a head of your competition.
