Is machine learning part of AI

What is the difference between AI and machine learning?

Artificial intelligence (AI) is on everyone's lips. Everyone knows what AI means. Or not? Finding a definition for artificial intelligence is not that easy. Mainly because many more are associated with this term. Related terms from the field of artificial intelligence are often equated with this. But that is not entirely true. We clarify the difference between artificial intelligence and machine learning.

AI and machine learning - relatives or just acquaintances?

The term artificial intelligence is now almost like a brand. Perhaps you are familiar with this phenomenon: handkerchiefs are generally known as “Tempo” and paper towels are known as “Zewa”. It doesn't matter which brand the products actually belong to. It is similar with the concept of artificial intelligence. Many use the term e.g. B. when they speak of humanoid robots from science fiction. Others refer to simple spreadsheets as artificial intelligence.

The reason for this vague use of the term is, on the one hand, the unclear definition. For years experts have been arguing about what “intelligence” is. So it is not surprising that the term “artificial intelligence” is also difficult to define. On the other hand, many do not understand which processes require intelligence and which do not. Climbing stairs is z. B. a complex sequence of movements that requires the cooperation of muscles, nerve tracts and the brain. For us humans, the process is self-evident and the many intermediate steps happen almost automatically. For computers, however, it is a difficult, complex task to learn. In contrast, arithmetic is easy for computers, while it often demands the greatest concentration from us humans.

All of these factors make it difficult to clearly define the concept of artificial intelligence. However, there are some indications. The two "A's" autonomy and adaptability - these keywords determine what artificial intelligence is and what is not. A computer is considered autonomous when it can perform complex tasks and does not need humans to do so. A computer is adaptable when it learns from experience and thus improves the result. This is where machine learning comes in. It is often mistakenly equated with artificial intelligence. In truth, machine learning is a branch of both artificial intelligence and statistics. Machine learning is one way that computers are adaptable. It works on the basis of experience or data. The more information a computer receives in the form of data, the better it learns to solve complex tasks.

An example: A system for autonomous driving should reliably detect traffic lights in road traffic. The more pictures it knows of traffic lights, the more reliably the system can identify every type of traffic light. It learns through a large number of traffic light images and becomes more adaptable and intelligent. Machine learning is therefore an important prerequisite for artificial intelligence.

AI has many faces - terms related to intelligent machines

In connection with artificial intelligence, you may have come across the terms “general” or “specific” artificial intelligence. You may also have heard of “strong” and “weak” artificial intelligence. We want to shed light on the darkness and create clarity in relation to the various “faces” of the AI.

General artificial intelligence concerns machines that can perform many different intellectual tasks. Specific artificial intelligence, on the other hand, describes computers that are specialized in one task. We find this form in our everyday life. Chatbots, driver assistance systems and music or film streaming services with personalized content are all specific forms of artificial intelligence.

In addition, there is a distinction between “strong” and “weak” artificial intelligence. The former describes a machine intelligence that can be equated with or even surpasses the human mind. The latter stands for systems that partially reach human intelligence, but are not intelligent in the human sense. Examples are image, speech and text recognition.

Intelligence vs. intelligent behavior - is there a difference?

To answer the question right away: the philosopher John Searle thinks yes. The famous thought experiment "The Chinese Room" comes from him. Imagine a person with no knowledge of Chinese being locked in a room. The person is given questions in Chinese through a doorway into the room. Using a manual, the person can seemingly answer the questions in Chinese, even though they do not speak the language. To outsiders, it appears that the person in the room has an excellent knowledge of Chinese. She looks intelligent. The fact is, however, that it only behaves intelligently. Although she answers the questions correctly using the manual (intelligent behavior), she cannot speak Chinese (intelligence).

The Chinese Room - or: Intelligence vs. Intelligent Behavior

John Searle transfers this thought experiment to the artificial intelligence of computers. A controversial question arises: are computers intelligent or do they just behave like that? According to Searle, computers cannot have the knowledge and understanding that humans have. Regardless of Searle's objection, however, we can state that the intelligence of machines is helpful in solving practical problems. Whether it is just about intelligent behavior or intelligence is of secondary importance for the time being.

Alan Turing and the Turing Test

The Turing Test, named after the pioneer of computer science, is better known than the Chinese Room. In the test, a person (in graphic C) exchanges written information with two players (in graphic A and B). If a person cannot determine whether A or B is the computer, i.e. the “intelligence” of a computer does not differ from that of a person, then the computer has passed the Turing test.

The Turing Test

 

Credits: Juan Alberto Sánchez Margallo / CC BY (https://creativecommons.org/licenses/by/2.5)

We state: machine learning is a branch of AI. With the help of large amounts of data, computers learn to be increasingly adaptable and autonomous when dealing with complex tasks. Artificial intelligence is characterized by these abilities. In this way, machines are able to carry out work processes automatically. A great help for us humans!

Machine learning and crowd guru - we train algorithms

Algorithms have to learn from a lot of data before they can act adaptively and autonomously. Our crowd takes on the algorithm training for your company. Since your order is being processed by hundreds of crowdworkers (our gurus) in parallel, you can count on quick results. So you have z. B. via a well-functioning chatbot in a significantly shorter time than if you learn the algorithm internally in your company.

Our gurus also have a wealth of experience in intelligent image segmentation for autonomous driving. Read more about it in our case study: Algorithm training from the crowd - intelligent image segmentation for autonomous driving.

If you are interested, please do not hesitate to contact us. We look forward to you!

 

Author: Katharina Strauss

Keywords: AI, AI, artificial intelligence, machine learning, machine learning