What is artificial intelligence (AI)? Basic Concepts of Artificial Intelligence

On this occasion we will learn What is artificial intelligence (AI) and Basic Concepts of Artificial Intelligence (AI), where many things will be learned from understanding to These products are grouped into There are four techniques in AI.

What is AI

Understanding AI from several sources can be known from 4 the opinion of the following experts:

• Automation of activities related to processes thinking, problem solving and learning (Bellman, 1978).
• The study of the ability to sense with using a computational model. (Charniak+McDermott, 1985).
• Study of how to do something so that it becomes better (Rich+Knight, 1991)
• The branch of computer science that focuses on automation
• smart behavior. (Luger + Stubblefield, 1993)

Broadly speaking, AI can be divided into 4 categories, namely:

• Thinking Humanly This approach is carried out in two ways: first, through introspection, trying to capture our thoughts ourselves when we think. “how do you know that you understand?”. The second is through research from a psychological point of view.
• Acting Humanly (The Turing test approach, 1950) This approach in 1950, Alan Turing designed a test for intelligent computers (Smart bots) to test whether the computer is able to trick a human/interrogator through communication remote text based. Of course the computer must have the ability, Natural Language Processing, Knowledge Representation, Automated Reasoning, Machine Learning, Computer Vision, Robotics.
• Thinking rationally In this approach there are two problems, namely: it is not easy to make informal knowledge, then express it in formal terms with logical notations. Which secondly there is a big difference between being able to solve problems “in principle” and solving them “in the real world”.
• Acting rationally (The Rational agent approach) This approach makes logical inferences which are part of a rational agent. Because to do action rationally is to reason logically, it can be concluded that the action taken will achieve the goal or not.

Until now, human thought that is beyond rationality, namely reflexes and intuitive (related to feelings) cannot yet fully imitated by computer. The two definitions above are considered inaccurate at this time. If you use this definition, then many current AI products that do not deserve to be called smart devices. The most appropriate definition of AI today is acting rationally.

AI Foundation

Humans are endowed with extraordinary intelligence. At the age of 3 years, he is already able to recognize various kinds of objects even though only partially visible. When he sees some of the lizard tails, then he will easily recognize that the animal is a lizard hiding behind a painting frame.

In adulthood, intelligence continues to grow rapidly, starting from cognitive, emotional and spiritual intelligence. So far, no machine has been able to equal human intelligence as a whole. For years, scientists have been trying to study human intelligence. 

From the thoughts of these scientists, then AI was born as a branch of science that seeks to understand human intelligence. Support for technological developments, both hardware and software very diverse software. So far, AI has produced There are many smart devices that are very useful for human life. Until now, AI continues to be studied and developed systematically broad and deep.

AI history

The term AI was first coined in 1956 in Darthmouth conference. The following is a brief summary of related stages of AI development history:

• The electronic computer era (1941) A tool has been found as an electronic computer that developed in the USA and Germany. The computer requires a spacious room and a separate AC room and involves the configuration of thousands of cables. This discovery form the basis for the development of programs that lead to AI.
• AI Preparation Period (1943-1956) Warren McCulloch & Walter Pitts managed to make a artificial neural cell model (1943). And Norbert Wiener made research on the principles of feedback theory (1950). Whereas John McCarthy (father of AI) conducts field research Automata, ANN and intelligence learning with make programs that can think.
• Early Development of AI (1952-1969) Newell and Simon's success with the "General" program Problem Solvers". This program is used to complete human problem. McCarthy demonstrates language high-level programming i.e. LISP at MIT AI Lab. Then Nathaniel Rochester of IBM and his students issued an AI program namely "Geometry Theorm Prover" which is able to prove a theorem (1959).
• AI development slows down (1966-1974) The emerging AI programs contain little or even contain no knowledge at all on the subject. The number of problems that must be solved by AI, because too many problems were related, it is not uncommon for failure to occur when create AI programs. There are some limitations on the structure basis used to produce behavior intelligence, for example, two different data inputs are not can be trained to recognize the two different inputs.
• Knowledge-based systems (1969-1979) Ed Feigenbaum, et al, developed a program to solve molecular structure problems (Dendral Programs) focused in terms of chemistry. And Saul Amarel in the project “Computer in Biomedicine” makes programs in terms of medical diagnostic knowledge.
• AI became an industry (1980-1988) When you become an AI industry, you find an expert system (R1) capable of configuring systems computer. And the AI ​​industry boom also involves a lot large companies that offer software tools for build an expert system.
• The Return of Artificial Neural Networks (1986-present) During this year Hopfield developed the technique statistical mechanics for optimizing artificial neural networks (1982). And David Rumelhart & Geoff Hinton discovered back-propagation algorithm. This algorithm works implemented in the fields of computer science and psychology (1985).

Current AI

Various AI products are successfully built and used in everyday life. These products are grouped into

There are four techniques in AI, namely:

• Search
• Reasoning
• Planning
• Robotics
• Learning

Examples of successful AI products built and deployed in everyday life, namely:

• MedicWare (Patient's medical record)
• Speech Processing (Voice Recognition, Speaker Recognition) 
• GPS (Optimal Route)

        - How can you find out someone's location (friends, other people) next time?

        - How the system can provide location recommendation which should be visited (eg going to campus) at a certain time?

• Catur

look at the game of chess

An example of a solution using AI in a Chess Game

• Super Mario

Example of a solution with take advantage of AI in Super Mario Games

• Game Dino AI

Solution using AI

        Information:

            Input = results from sensors or observations around agent

            Yellow color = distance from obstacle

            Red color = obstacle height

            Blue color = obstacle speed

            Output = can be automatically action

• Computer Vision

What Computer Vision includes

Use of AI for face detection

Use of AI to calculate Number of Vehicles

Use of AI to count Number of People

• Optimization = In optimization techniques can use several algorithms, among others;

                                    - Genetic Algorithm (GA)

                                    - Particle Swarm Optimization (PSO) Algorithm

            Examples of optimization problems include:

1. Problems in scheduling lectures, problems which can be taken are:

- Day selection: each lecturer can set the day and hour usually in teaching.

- Selection of SKS & MK: each lecturer can set the number of credits & MK minimum and maximum that is in the amp.

- Output: optimal class schedule

Use of Optimization Algorithms

Results from leveraging AI

2. The problem of scheduling a doctor's watch, a problem that can be taken namely;

- Day, every doctor can set the usual day and time teach

- Places of guard, for example 2 places, in hospitals and clinics

- Output, optimal schedule for hospitals and clinics

Doctor Watch Scheduling Optimization:

AI of the Future

In 2009, a $1000 PC will be able to do about one trillion calculations per second and equivalent to ability computational human brain consisting of Virtual Reality and computer interaction with body cues. The year 2029 is also a possibility a $1000 PC will equal the computational power a thousand human brains consisting of computers can be connected with a human brain and a computer capable of reading all literature and multimedia materials. In 2049, food is produced using nanotechnology. Then in 2072, the possibility of picoengineering or technology on the picometer or 10-12 . scale meter was successfully applied. Then in 2099, there may be a tendency to make a combination of thoughts human and machine intelligence. So we can no longer distinguish whether the agent is a machine or a human.

Simple Case Study (Logic 1)

Examples of problems in the game below:

Game View

See : Sorts of Machine Learning Algorithms You Should Know

Resource :  

Cholissodin, I., Sutrisno, S., Soebroto, AA., Hasanah, U., Febiola, YI., 2020, AI, Machine Learning & Deep Learning, Fakultas Ilmu Komputer, Universitas Brawijaya, Malang.