Today, one of the most prominent fields of study within artificial intelligence is machine learning. As discussed earlier, artificial intelligence is broadly defined as computer systems capable of performing intelligent human-like processes. Also previously mentioned, the greater scope of AI includes both narrow and general systems. For the most part, the technologies being utilized now fall under the classification of narrow AI. This is due to the fact that they are meant to operate in very specific situation and only need to perform predetermined tasks.
Machine Learning – Photo Credit
Meanwhile, machine learning systems are much closer to the classification of general intelligence. This is due to the fact that narrow AI is only able to learn and develop in a certain area, while machine learning is defined “as an application of artificial intelligence that provides systems the ability to automatically learn and improve from experience without being explicitly programmed.” They may not be quite worthy of the classification general AI, but they are certainly closer then the narrow AI systems be implemented today.
Essentially, the goal of machine learning is to develop systems that can learn like humans. This takes it a step further than just completing human-like process. Narrow AI systems have expectations and predefined code to produce specific output and receive certain input, meaning only the values are changing. Most AI systems fall into this category as they are built with sufficient base code that is relevant to the tasks being asked of it. Machine learning take it a step further as these systems are not prompted to receive the input and commands, and they are not prepared for the outputs they will produce.
It is the differentiation of how the different systems came to be that separates them. Most AI systems are usually classified as narrow and have significant base code that is relevant to the tasks being asked of it, meaning there are already components that take input and produce output. Within a narrow system the algorithm is only modified to make the outputs increasingly more accurate as the system is trained. With machine learning lacked, all of these processes are developed over time as they originate with nothing but the ability to learn.
Machine Learning Web – Photo Credit
Well then, how do these systems learn? Simply put, it is fed a significant amount of data and creates a model from this data. So, when it is first fed information it finds patterns within the data and creates a model. Once that model is created it attempts to make predictions based on the model based on inputs. This output is then put through a learner which assesses the accuracy of the output by some manner. The learner can be either supervised or unsupervised, which won’t be touched on today, but is in the graphic above. Once this is done, the parameters of the model are modified to make the model more accurate as a whole so that it will be able to make better predictions in the future. This system differs from narrow AI systems in that the basics of the model would be initially provided, giving the technology the fundamentals before training it.
Machine Learning Explanation by Google – Photo Credit
Machine learning may seem unnecessary when looking at it compared to other forms of artificial intelligence that have frameworks related to what they are doing. However, by setting these systems free with no basis they can be applied on larger scales and one system can be implemented in a variety of ways. Today, machine learning is being used for visual recognition, speech recognition, medical diagnoses, statistical predictions and much more. It is difficult to determine the limits of this technology today and it will be interesting to say what it will accomplish.