Machine learning and artificial intelligence are also frequently used words in our daily lives, such as deep learning and artificial neural networks, which leads to confusion. In today's article, we will address the differences between these terms and eliminate some of the term confusion in the middle.
Machine learning is a sub-branch of artificial intelligence and deep learning is a sub-branch of machine learning. Artificial neural networks form the basis of deep learning.
What are Artificial Neural Networks?
We answer this question in detail in our article "What are Artificial Neural Networks?". I suggest you read this before continuing with our article. However, to put it simply, artificial neural networks are the result of mathematical modeling of the human brain. It is an algorithm inspired by the working structure of the human brain.
What is Deep Learning?
Counting an artificial neural network as a deep learning algorithm depends on the number of hidden layers used, and this value should be 3 or more.
Differences in Deep Learning and Machine Learning
Deep learning is a sub-branch of machine learning. The biggest difference between the two is how algorithms perform the learning process. The process of elimination of arguments that will not be used in deep learning occurs automatically. In classical machine learning algorithms, however, such as linear regression or logistical regression, people study the connections between arguments and decide whether they are suitable for the problem and then give the algorithm the appropriate arguments. So classical machine learning algorithms don't have human intervention, but not in deep learning algorithms. We give everything we have as data on the problem in deep learning to our model and our model will do the rest.
Deep learning algorithms achieve appropriate results by identifying the major patterns on their data. Therefore, it needs much more data than conventional machine learning models to achieve better success on the given problem. However, according to machine learning models, it gives more successful results in more complex problems. Examples of these problems include object detection, fraud detection, or virtual assistants.
Another important difference is that deep learning algorithms are much larger than classical machine learning algorithms and also have more parameters, which requires very serious processing power during training and operation. GpUs are used instead of CPUs to meet this processing power, and parallel programming is used. Therefore, while it is not a very practical machine learning algorithm, with the development of GPUs and easier access to GPUs, there have been rapid improvements in the field of deep learning and its popularity has increased.
What is Artificial Intelligence?
AI is the most wide range of terms we mentioned and covers all of them. There are 3 main categories of artificial intelligence:
- Artificial Narrow Intelligence
- Artificial General Intelligence
- Artificial Super Intelligence
Narrow AI is also defined as "weak" artificial intelligence. The other two categories are defined as "powerful" artificial intelligence. Weak AI focuses on solving specific problems. Examples include playing chess or getting to know a specific person in photos.
General AI is a category where human traits such as understanding emotions and tones outweigh. Although chatbots and virtual assistants (Siri, for example) are close to the overall AI category, they are still classified as weak AI.
Super AI is aimed at the human brain. Although there is no example of the strong AI category today, studies in these two categories (general and super) are ongoing. An example of what awaits us in these two categories in the future is dolores in Westworld.