Blog: A simple way to explain the key differences between Artificial Intelligence, Machine Learning &…
Artificial Intelligence (AI) and Machine Learning (ML) are often used interchangeably. Deep Learning (DL) often regarded as an equal term as ML. However, they are distinct in their practice as well as the underlying technology employed.
AI tends to be the overarching term for many layers of practices or methods, but it is the theory and development of practices which enable machines to perform tasks that are characteristic or mimic of human intelligence.
These intelligence or smart tasks are usually the processing of unstructured data, such as vision and speech recognition, decision making, and language translation.
ML, however, is simply a means to achieve AI efficiently. Through ML, a machine is trained with a large amount of data and algorithms which allow it to perform specific tasks. For instance, a machine could be trained to distinguish between cats and dogs by reviewing hundreds, or even millions of pictures previously tagged and categorize by people.
Deep Learning (DL) is also a means of achieving AI and functions very similarly to ML. However, it takes the learning process a step further. For example, a machine trained to identify images of dogs and cats through DL may begin to recognize that an image of a lion or wolf could fit those categorizes as well despite not having trained with this information.
The underlying difference between ML and DL is the Artificial Neural Networks (ANN), which DL uses it to learn and make decisions. These neural networks are based on the modern understanding of how human brains work with high levels of interactions, interconnection, and neural activity.
Similar to a human brain, which it recognizes errors and makes informed decisions in new environments based on past experiences, ANN create a system of probability making decisions with a degree of certainty based on the data that it learns from. Also, there is a feedback loop that allows the network to learn more and refines its techniques which maximize its accuracy. This ability to respond to new information, learning, and responding continuously is a critical advantage of this technique over other standard or deterministic techniques. This is a very useful feature when analyzing information that is unexpected and nonlinear.
To summarise, DL is a subset of ML, which both are subsets of and means of achieving AI. Real life examples to contextualize the summary.
AI-based machine: Deep Blue that beat Garry Kasparov 20 years ago.
ML: Email Spam filter that improves along the way as it receives more inputs in the form of spam.
DL: Computer vision, in particular by Google, can recognize images by itself.
In the next writing, we shall briefly touch on how a machine learns.