Original article was published on Artificial Intelligence on Medium
Artificial intelligence is the new electricity — Andrew Ng
Electricity has changed how the world operated. It changes transportation, manufacturing, agriculture, and even health care. For example, before the invention of electric lighting, humans were limited to daytime activities, because at night it was dark, only people who could afford gas lamps could do activities. Compared to now, we can still do activities at night because it is illuminated by electric lights.
Likewise with Artificial Intelligence (AI). AI is expected to have a similar effect. A concrete example that we can take today is the advertising problem. Previously, the ads that appeared were determined manually by humans. Now, advertisements that appear are determined by AI, based on the user profile data. For example, if the user likes fashion then what might appear is a clothing advertisement.
As technology advances, research in the field of AI is also growing, so the previous benchmarks that define Artificial Intelligence (AI) are now becoming obsolete.
From this development came new terms such as machine learning and deep learning. But sometimes, there are slices between AI, machine learning, and deep learning, so the difference between them can be very unclear. So in this post, I’ll give a quick explanation of what AI, machine learning, and deep learning mean and how they are different.
Let’s start with this image.
When viewed from the picture we get a little picture that machine learning is part of AI and deep learning is part of machine learning. However, some people argue that AI and machine learning are separate. However, in this article, I will be in the argument that machine learning is part of AI. For more details, let’s be more clear about what is AI.
AI has been studied for decades and is still one of the most difficult subjects to understand in Computer Science. This is partly because of how large and vague the subject is. This has applications in almost every way we use computers in society. AI is anything that refers to the Simulation of Human Intelligence in machines that are programmed to think like humans and imitate their actions.
The term artificial intelligence was first coined by John McCarthy in 1956 when he held his first academic conference on the subject. Although, the journey to understanding if a machine can really think begins long before that.
AI typically analyzes its environment and takes actions that maximize its chances of success. In early days, approaches to AI are such as formal logic and expert systems. These methods dominated AI at the time.
However, as the development of computational power, greater emphasis on solving specific problems, and also there are new ties between AI and other fields. And one method that rises by the effect of this is learning or we will called it machine learning.
In accordance with the words, machine learning means machines that learn from data. Machine Learning mimics the concept of human learning from childhood to adulthood. For example, if a baby is taught to distinguish between dogs and cats, they will be given pictures of dogs and cats, as well as machines, if you want to make a machine learning we must provide knowledge to machines about dogs and cats. And the more pictures that are given the smarter the machine is to distinguish between cats and dogs.
Machine learning is closely related to computational statistics, which focus on making predictions. Data mining is also related to this study, which focuses on exploratory data analysis through unsupervised learning.
In machine learning, there are several types of algorithms used and are grouped based on the expected input and output of the algorithm.
Supervised learning create functions that map an input to the desired output, for example in classification. It observed patterns of data and converts them into model to predict future data. For more details, we see this example.
It can be seen from the illustration above that at first we only had data with the cross and round classes. Then we have new data, triangle. The new data is then predicted based on models that have been previously trained. Because the new data has similarities to the cross classes, these data are classified into cross classes.
An example of a method that is included in supervised learning is neural network, kNN, decision trees, naïve bayes, SVM, etc.
Unsupervised learning models the input set, such as clustering. Unlike the classification that each data has a class. Clustering works by grouping similar data.
From the illustration above it can be seen that at first, we have a set of data, then the data is grouped into certain groups based on the similarity of the data with other data.
An example of a method that is included in supervised learning is k-means, DBSCAN, etc.
Reinforcement learning is a learning algorithm that is applied to intelligent agents so that they can adjust to the conditions in their environment, this is achieved by maximizing the value of the ‘reward’ prize that can be achieved. This type teaches how to act to deal with a problem, an action that has an impact.
The most common example is a self-driving car, the machine is asked to avoid collisions or violations. If an accident or violation occurs, then the machine will be given a negative reward, if the machine does it right, it will be given a positive value. From there the machine will learn to drive a car.