The first half of the above video reflects a typical impression of artificial intelligence as something very powerful, difficult to understand and especially difficult to control once it has taken over.
It is no wonder. Science fiction has influenced people’s imaginations and for the most of us, there exists no counterbalance to that image in our daily life. Take, for example, T-800, the human like character of the movie “Terminator” portrayed by Arnold Schwartzenegger (
However, this futuristic image of our Doomsday clip above, as well as the mentioned movies, has a little if anything to do with the current development of AI and its practical applications. Even though we already have, for example, self-driving cars (e.g.
This all means also that AI is not yet (if ever will) capable of generating more AI: the practical applications are now (and in the future) done and used by us humans. And it requires a lot of people in many different levels. People that are able to implement the AI. People that are able to innovate new applications. People that are able to describe these visions. People that are able to explain the nature of the application fields to all these people. And people who just understand the possibilities and principles of AI.
Would you like to be one of them? If yes, this course could be your first step on that path.
If you are seeking for a definition of “Artificial Intelligence” and perform a web search, you might notice that this term has a different meaning to different people. Some might say that AI is an artificial life-form that is able to surpass human intelligence, and for others, AI could be any technology that processes data. Thus, before going more deeply into that, let us see what we can extract from its components, the words “Artificial” and “Intelligence”, separately.
The word “Artificial” might mean something fake or forced such as in “artificial smile” or “artificial flavor”. But in many contexts, it refers to something that imitates a natural thing but is made by us humans. This is roughly also the meaning with the context of AI.
For the word “Intelligence” the definitions are a bit more complex and variable. By intuition, all but the simplest human behavior is related to intelligence, while complicated insect behaviors would not be considered intelligent. You may also recall a term intelligence quotient, IQ, that is being used to measure some human abilities that reflect some specific fields of cognitive skills. In human psychology human intelligence profile is often determined by Wechsler Intelligence Scale Test, a set of standardized tests that measure abilities in many different fields that are necessary in human life. Theses tests each look at human abilities in narrow fields such as vocabulary of a person or arithmetic skills.
by
Ants possess different and distinct modules in their brains. These modules are dedicated to different navigational tasks. For example, one module keeps track of distance traveled. A second module is dedicated to visual scenery, it allows ants to recognize important routes. This modular structure of ant brains is a brilliant example of an intelligent system where different modules are focusing on a narrow specific problem. The information from these modules being combined makes the whole system, an ant, more capable of solving different problems.
Despite this, a simple individual ant could not be considered as very bright. It is not able to alone survive in harsh conditions nor can it perform complex actions. While each ant is focusing on one single task only at a time, different ants in a colony may perform different tasks. Moreover, the ants can change their behavior and even the focus tasks based on the chemical information they get from their fellow ants. This makes the whole colony able to adjust with changing environment even if no single ant is giving any orders. Single ants work in roles of sensors making simple observations on their path and informing other ants of those observations by leaving chemical traces. And while this information sweeps through the colony, the whole system is able to adapt itself to the changed situation.
So, an ant colony can find solutions to problems by changing the behavior of individuals and thus the behavior of the entire colony. But can this behavior be considered as intelligent? What intelligence is?
Giving a single valid definition for the term “intelligence” is not possible. Wikipedia alone (
This all is true with an ant colony in small scale and with the human life in large scale. But does it also hold in the context of “Artificial Intelligence”? And how to know if a machine is intelligent? We will next introduce two famous approaches, the Turing test and the Chinese room experiment, that aim to enlighten this question a bit.
An English mathematician Alan Turing (1912 - 1954) is often named as the founder of computer science. Among many other achievements, he was in an essential role in designing and constructing the very first computers ever built. Honoring his achievements his picture was printed on the £50 note of Bank of England in June 2021.
Alan Turing wanted to be able to identify if a machine was equally intelligent as a human. For this, he described an imitation game where a true person, a human evaluator, would be communicating with two other parties: another human and a computer. The communication would be done only in a format that would not reveal the nature of the party, for example in text format. The aim of the game for the evaluator would be to distinguish which one of the other parties is a computer and which one is human. If the computer would be evaluated as a human, it would be considered as intelligent.
A computer that would pass the Turing test would give an impression of being able to think by itself. But could it really think? This question is in behind another famous thought experiment made by John Searle (1932- ) that aims to illustrate the nature of machine intelligence: the Chinese room experiment. Mr. Searle himself did not know any of the Chinese language nor the Chinese characters. He pictured a situation where he would be inside a room with only small openings to receive and output notes written on paper.
From one opening he would receive notes that have phrases written in Chinese. He would use books or a computer program to interpret the Chinese characters into English, respond to the phrase and then use the books again in mechanically translating the response into Chinese characters. For the ones outside of the room, feeding in the papers and receiving the responses, this would look like the person inside understands Chinese and they would also see him passing the Turing test. However, as we know, there would be no understanding at all and thus no thinking or reasoning.
The point of this experiment is that what happens with the man in the Chinese room is what happens with computer programs as well. They both follow predetermined rules in producing solution to problems, but they are not actually able to understand or give meaning to the communicated phrases. Merely it is about simulating the understanding of Chinese. This experiment describes well the level of Artificial Intelligence as we face it today and most probably also far in the future.
We used the terms Strong AI and Weak AI already a couple of times and now it is a good moment to get back to them shortly.
Looking at the Chinese room experiment, what happens when the person transcribes the Chinese characters and mechanically produces a response looking valid for the persons outside, could be seen as Weak AI. It applies some predefined rules for solving a problem for which it has been prepared. If the person inside would not need any tools but were able to produce the output by himself, that would represent some Strong AI. It would no longer be a simulation of a mind but having a mind.
Regarding the Artificial Intelligence, we are still far away from systems that could be considered as Strong AI.
So, now that we know roughly what the word “Intelligence” means, we should be able to define the term “Artificial Intelligence” as well.
Apparently, it refers to something made by us humans that has or at least imitates a capability of, for example, understanding, reasoning, adapting to new conditions and learning from experience and, by doing so, providing solutions to problems. This capability can have a narrow focus, it may be specific to a limited field of problems or even only to a subset of such a field or, it may as well have a more general focus.
And this kind of capability can be implemented in many ways. Below you find a list of few branches of AI explained shortly, each one of them representing one specific strategy or field of strategies for the implementation of AI (Source: Prateek Joshi 2017, Artificial Intelligence with Python).
It is not required or even practical to try to map and memorize all branches of AI and their definitions. The key thing to remember is that AI includes a wide range of approaches. It comes in many forms and shapes and appears in many kinds of applications such as computer vision, natural language processing, and robotics.
By now, you should have some idea of what AI is and what kind of problems can be solved using AI and machine learning. So far, many examples have been trivial and not easily understood in real life. Let's take a look at what industries are, for example, using AI as of today.
Below is a list of industries in no particular order:
Or let us look at the different fields of retail business. Even a small brick and mortar business might apply machine vision and machine learning for finding a layout for the shop that would improve the profitability. It might also be predicting the daily demand of different products to continuously keep the shelves optimally filled and cargo costs as low as possible. And if the shop would also sell its products in web, it could monitor the behavior of the customers in the web shop and, by that data, optimize the web shop structure, product offer and the prices.
For each question, pick up the choice that best answers the question, fills up the sentence or defines the concept