Table of Contents
Artificial intelligence, machine learning, and deep learning are now an integral part of many enterprises. At the same time, artificial intelligence takes huge steps towards progression. It has come a long way from achievements in the field of unmanned vehicles and the ability to beat a person in games such as poker and Go, to automated customer service.
Often the terms artificial intelligence, machine learning, and deep learning are used haphazardly as interchangeable, but, in fact, there are differences between them. Let’s understand what all the noise is about and what potential benefits they can bring to business.
Machine Learning vs Deep Learning: What is the Difference?
Machine learning is the process of artificial intelligence implementation.
In 1959, Arthur Samuel coined the term "machine learning" - the ability of a computer to learn without human intervention. Artificial intelligence is possible without machine learning, but this will require a million lines of code with complex rules and conditions.
In other words, instead of writing down detailed instructions for each specific task, an algorithm is used that learns to find solutions on its own. First you collect hundreds of thousands of pictures, and then mark on them some cats or whatever. The algorithm builds a model for the computer so that it can identify and highlight pictures of cats as a human does. As a result, it “learns” what cats look like and becomes able to recognize them with a minimum error rate.
Deep learning, in its turn, is just one of many machine learning approaches.
Examples of other approaches: decision tree analysis, inductive logical programming, clustering, reinforcement learning, Bayesian network.
Deep learning is inspired by the structure of the human brain and the interaction of neurons. Involved algorithms imitate the organization of the brain creating an artificial neural network (ANN).
ANN consists of artificial neurons that interact with each other. They are arranged in layers - each layer reacts to certain signs, for example borders of figures during image recognition. Because of multilayered structure it is called deep learning.
To put it simple, deep learning is the use of neural networks with a large number of neurons, layers and relationships. We are still far from imitating the human brain in all its complexity, but we are already moving in that direction.
To summarize. Machine learning is a subset of artificial intelligence associated with the creation of algorithms that can change themselves without human intervention to obtain the desired result. To achieve it you have to “feed” them clear and structured data. For machine learning to work as intended, there have to be an array of structured information. Deep learning is less demanding on data structure, and therefore can interact with photos, video and sound, revealing patterns in them.
So what is the practical use of machine learning?
We’re already surrounded by machine learning. It is used in the programs installed on our smartphones, in cars and smart homes. Moreover, it is already present in the software that we use at work to analyze information and make data driven decisions in less time. According to Gartner analysts, over the past year, the dynamics of artificial intelligence technologies implementation increased from 4% to 14%.
Machine learning possibilities are almost endless. It can be used wherever fast data analysis is important, and can have a revolutionary effect where it is important to identify trends or anomalies in large data sets. From clinical trials to the field of safety and monitoring compliance with standards.
For example, American Express processes transactions worth $1 trillion. They rely heavily on data analytics and machine learning algorithms that help detect fraud in real time, thereby saving millions of dollars in loss. Additionally, AmEx uses its data streams to develop applications that can associate the cardholder with products or services and special offers. They also provide online sellers with business trend analysis and industry benchmarking.
Healthcare is also actively introducing ML. As of 2019, machine learning algorithms are even capable of detecting cancerous tumors and skin cancer, diagnosing diabetes and, most importantly, predicting the progression of the disease. It’s not a surprise, that by 2025, the medical AI market is expected to grow to $34 billion.
There is another good example. Social network Pinterest is already implemented machine learning to improve user experience. For this purpose, Kosei, a machine learning company, was acquired. Today, machine learning is involved in every aspect of Pinterest’s business operations, from moderation of spam and content searches to monetizing ads and reducing the number of unsubscribes from newsletters. Thus, the social network provides a much more convenient search, user interaction and facilitates the work of moderators, increasing its profit.
Machine learning is rapidly becoming the driving force behind the revolution in customer service. As a rule, a huge array of requests from consumers can be divided into a few categories, and many of them are easily predicted. Chat bots based on machine learning may well respond to them. This reduces waiting time and the number of dissatisfied clients, which makes the business more efficient. It also allows customer service managers to spend their time only processing unique complaints and requests that really require human intervention.
What about deep learning?
Most of the projects with deep learning are used in the recognition of photos or audio area, or diagnosis of diseases. For example, it is already used in Google translations from the image: Deep Learning technology allows you to determine if there are letters in the picture, and then translate them. Another project that works with photos is a face recognition system called DeepFace. It can recognize human faces with a 97.25% accuracy rate.
At the same time, deep learning can semantically segment an image or video - that is, not only indicate that there is an object in the picture, but also perfectly highlight its shape. This technology is used in driverless vehicles to detect the presence of interference on the road, marking and read information from road signs to avoid accidents. A neural network is also used in medicine - to determine diabetic retinopathy from photographs of patients' eyes, for example. The U.S. Department of Health has already authorized the use of this technology in government clinics.
However, the scope of deep learning is not limited to working with photos, video and sound. For example, recently, a new neural network-based Google learning algorithm has learned to optimize the placement of components on a computer chip to make it more efficient and less energy-intensive.
The implementation of this technology began relatively recently, in the previous decade. Before it was expensive, difficult and time-consuming process. Powerful graphics processors, video cards and memory were needed. The boom in deep learning is precisely connected with the widespread use of graphic processors that speed up and reduce the cost of computing, virtually unlimited data storage capabilities and the development of "big data" technology.
It’s hard to say whether this technology will be disruptive. On one hand, Google, Facebook and other large companies have already invested billions of dollars and are optimistic about it. In their opinion, deep-learning neural networks are capable of changing the technological structure of the world. On the other hand, there are skeptics: they believe that deep learning is a buzzword or rebranding of neural networks.
So what to choose?
It depends on what tasks you are going to perform using AI. If you need to process a large amount of structured data to get some kind of conclusion, machine learning is the best solution. If we are talking about the tasks associated with processing an array of graphic images, sound, video, then it is better to apply deep learning.
Mentally, you are probably already flipping through the catalog of droids, but return to reality and first figure out how to get the most benefit from this technology for yourself. We are fortunate to live in a time when we can already use advanced technologies of today to grow our business. And as you think about the next step, we're taking all the efforts for its implementation.