Elements of AI Part 1 Chapter 1. What is AI?
Section 1. How should we define AI?
Link - https://course.elementsofai.com/1/1
Let's discuss the three main applications of AI to demonstrate the different aspects of AI.
Application 1. Self-driving cars
Self-driving cars require a combination of AI techniques of many kinds: search and planning to find the most convenient route from A to B, computer vision to identify obstacles, and decision-making under uncertainty to cope with the complex and dynamic environment. Each of these must work with almost flawless precision to avoid accidents.
The same technologies are also used in other autonomous systems such as delivery robots, flying drones, and autonomous ships.
Application 2. Content recommendation
A lot of the information that we encounter in the course of a typical day is personalized. Examples include Facebook, Twitter, Instagram, and other social media content; online advertisements; music recommendations on Spotify; movie recommendations on Netflix, HBO, and other streaming services. Many online publishers such as newspapers’ and broadcasting companies’ websites as well as search engines such as Google also personalize the content they offer.
While the frontpage of the printed version of the New York Times or China Daily is the same for all readers, the frontpage of the online version is different for each user. The algorithms that determine the content that you see are based on AI.
Application 3. Image and video processing
Face recognition is already a commodity used in many customer, business, and government applications such as organizing your photos according to people, automatic tagging on social media, and passport control. Similar techniques can be used to recognize other cars and obstacles around an autonomous car, or to estimate wildfire population, just to name a few examples.
AI can also be used to generate or alter visual content. Examples already in use today include style transfer, by which you can adapt your personal photos to look like they were painted by Vincent van Gogh, and computer generated characters in motion pictures such as Avatar, the Lord of the Rings, and popular Pixar animations where the animated characters replicate gestures made by real human actors.
Reason 1: no officially agreed definition
Reason 2: the legacy of science fiction
Reason 3: what seems easy is actually hard and what seems hard is actually easy
“AI” is not a countable noun
When discussing AI, we would like to discourage the use of AI as a countable noun: one AI, two AIs, and so on. AI is a scientific discipline, like mathematics or biology. This means that AI is a collection of concepts, problems, and methods for solving them.
Because AI is a discipline, you shouldn’t say “an AI”, just like we don’t say “a biology”. This point should also be quite clear when you try saying something like “we need more artificial intelligences.” That just sounds wrong, doesn’t it? (It does to us.)
The use of AI as a countable noun is of course not a big deal if what is being said otherwise makes sense, but if you’d like to talk like a pro, avoid saying "an AI", and instead say "an AI method".
Exercise 1: Is this AI or not?
Section 2. Related fields
Machine learning can be said to be a subfield of AI, which itself is a subfield of computer science (such categories are often somewhat imprecise and some parts of machine learning could be equally well or better belong to statistics). Machine learning enables AI solutions that are adaptive. A concise definition can be given as follows:
Machine learning - Systems that improve their performance in a given task with more and more experience or data.
Deep learning is a subfield of machine learning, which itself is a subfield of AI, which itself is a subfield of computer science. Deep learning refers to the complexity of a mathematical model, and that the increased computing power of modern computers has allowed researchers to increase this complexity to reach levels that appear not only quantitatively but also qualitatively different from before. As you notice, science often involves a number of progressively more special subfields, subfields of subfields, and so on. This enables researchers to zoom into a particular topic so that it is possible to catch up with the ever increasing amount of knowledge accrued over the years, and produce new knowledge on the topic — or sometimes, correct earlier knowledge to be more accurate.
Data science is a recent umbrella term (term that covers several subdisciplines) that includes machine learning and statistics, certain aspects of computer science including algorithms, data storage, and web application development. Data science is also a practical discipline that requires understanding of the domain in which it is applied in, for example, business or science: its purpose (what "added value" means), basic assumptions, and constraints. Data science solutions often involve at least a pinch of AI (but usually not as much as one would expect from the headlines).
Robotics means building and programming robots so that they can operate in complex, real-world scenarios. In a way, robotics is the ultimate challenge of AI since it requires a combination of virtually all areas of AI. For example:
- Computer vision and speech recognition for sensing the environment
- Natural language processing, information retrieval, and reasoning under uncertainty for processing instructions and predicting consequences of potential actions
- Cognitive modeling and affective computing (systems that respond to expressions of human feelings or that mimic feelings) for interacting and working together with humans
Many of the robotics-related AI problems are best approached by machine learning, which makes machine learning a central branch of AI for robotics.
Exercise 2: Taxonomy of AI
A taxonomy is a scheme for classifying many things that may be special cases of one another. We have explained the relationships between a number of disciplines or fields and pointed out, for example, that machine learning is usually considered to be a subfield of AI.
Your task: Construct a taxonomy in the Euler diagram example given below shows the relationships between the following things: AI, machine learning, computer science, data science, and deep learning.
Exercise 3: Examples of tasks
Consider the following example tasks. Try to determine which AI-related fields are involved in them. Select all that apply. (Hint: machine learning involves almost always some kind of statistics).
Note: This exercise is meant to inspire you to think about the different aspects of AI and its role in various applications. As there are no clear-cut answers to many of these questions, this exercise will not be included in the grading. Nevertheless, we suggest that you do your best and try to answer as well as you can, but don't worry if our answers differ.
Section 3. Philosophy of AI
The very nature of the term “artificial intelligence” brings up philosophical questions whether intelligent behavior implies or requires the existence of a mind, and to what extent is consciousness replicable as computation.
The Turing test
Alan Turing (1912-1954) was an English mathematician and logician. He is rightfully considered to be the father of computer science. Turing was fascinated by intelligence and thinking, and the possibility of simulating them by machines. Turing’s most prominent contribution to AI is his imitation game, later known as the Turing Test
In the test, a human interrogator interacts with two players, A and B, by exchanging written messages (in a chat). If the interrogator cannot determine which player, A or B, is a computer and which is a human, the computer is said to pass the test. The argument is that if a computer is indistinguishable from a human in a general natural language conversation, then it must have reached human-level intelligence.
What Turing meant by the test is very much similar to the aphorism by Forrest Gump: “stupid is as stupid does”. Turing’s version would be “intelligent is as intelligent says”. In other words, an entity is intelligent if it cannot be distinguished from another intelligent entity by observing its behavior. Turing just constrained the set of behaviors into discussion so that the interrogator can’t base her or his decision on appearances.
The Chinese room argument
The idea that intelligence is the same as intelligent behavior has been challenged by some. The best known counter-argument is John Searle’s Chinese Room thought experiment. Searle describes an experiment where a person who doesn’t know Chinese is locked in a room. Outside the room is a person who can slip notes written in Chinese inside the room through a mail slot. The person inside the room is given a big manual where she can find detailed instructions for responding to the notes she receives from the outside.
Searle argued that even if the person outside the room gets the impression that he is in a conversation with another Chinese-speaking person, the person inside the room does not understand Chinese. Likewise, his argument continues, even if a machine behaves in an intelligent manner, for example, by passing the Turing test, it doesn’t follow that it is intelligent or that it has a “mind” in the way that a human has. The word “intelligent” can also be replaced by the word “conscious” and a similar argument can be made.
Is a self-driving car intelligent?
The Chinese Room argument goes against the notion that intelligence can be broken down into small mechanical instructions that can be automated.
A self-driving car is an example of an element of intelligence (driving a car) that can be automated. The Chinese Room argument suggests that this, however, isn’t really intelligent thinking: it just looks like it. Going back to the above discussion on “suitcase words”, the AI system in the car doesn’t see or understand its environment, and it doesn’t know how to drive safely, in the way a human being sees, understands, and knows. According to Searle this means that the intelligent behavior of the system is fundamentally different from actually being intelligent.
How much does philosophy matter in practice?
The definition of intelligence, natural or artificial, and consciousness appears to be extremely evasive and leads to apparently never-ending discourse. In intellectual company, this discussion can be quite enjoyable (in the absence of suitable company, books such as The Mind’s I by Hofstadter and Dennett can offer stimulation).
However, as John McCarthy pointed out, the philosophy of AI is “unlikely to have any more effect on the practice of AI research than philosophy of science generally has on the practice of science.” Thus, we’ll continue investigating systems that are helpful in solving practical problems without asking too much whether they are intelligent or just behave as if they were.
General vs narrow AI
When reading the news, you might see the terms “general” and “narrow” AI. So what do these mean? Narrow AI refers to AI that handles one task. General AI, or Artificial General Intelligence (AGI) refers to a machine that can handle any intellectual task. All the AI methods we use today fall under narrow AI, with general AI being in the realm of science fiction. In fact, the ideal of AGI has been all but abandoned by the AI researchers because of lack of progress towards it in more than 50 years despite all the effort. In contrast, narrow AI makes progress in leaps and bounds.
Strong vs weak AI
A related dichotomy is “strong” and “weak” AI. This boils down to the above philosophical distinction between being intelligent and acting intelligently, which was emphasized by Searle. Strong AI would amount to a “mind” that is genuinely intelligent and self-conscious. Weak AI is what we actually have, namely systems that exhibit intelligent behaviors despite being “mere“ computers.