DIU Blog

Artificial Intelligence – A Definition for Non-Techies

Artificial intelligence (AI) is the imitation of human intelligence. Unlike human brains, this intelligence is processed by machines, especially computer systems. The computers use three steps to process such intelligence.
They are:

1. Learning,
2. Reasoning and
3. Self-correction.

Let’s learn a little more about these three phases. In the learning phase, the machine collects information and uses a set of rules to use the information. In the reasoning phase, the computer reaches an approximate solution based on the rules established in the learning phase. In the self-correction phase, the computer compares its solution with the expected answer and modifies the process to produce a more accurate solution. And this is how a computer simulates the human brain. For Example – speech recognition systems, human face recognition systems, self-driving cars, etc.

We can categorize artificial intelligence (AI) as either weak or strong AI. The weak AI (or narrow AI) is designed and trained for a particular task. For Example, Apple’s Siri, virtual personal assistants, etc. Strong AI is known as artificial general intelligence. A strong artificial intelligence system has cognitive abilities like humans. That means they can find solutions in an unfamiliar environment without human intervention.

Artificial intelligence’s hardware, software, and management costs can be very expensive. That is why ‘Artificial Intelligence as a Service (AIaaS)’ is gaining popularity daily. Many vendors are offering cloud AI services. A few popular AIaaS providers are Amazon AI services, Google AI services, Microsoft Cognitive Services, and IBM Watson Assistant. Individual companies use these AI services for financial analysis, market demand forecast, intelligent demographic marketing, production management, etc.

Proper application of artificial intelligence in business makes it more profitable. It adds new functionality to business and helps deliver better service at a lower cost. However, sometimes the use of artificial intelligence raises ethical questions. Because the performance of AI systems depends on the training data. And training data are selected by a human. If the AI developers want, they can train their system to steal users’ information and influence them unethically. There is also a growing concern about the impact of artificial intelligence on the job market. However, some industry experts believe the fear is not realistic. The application of AI improves the quality of products and services. It creates job opportunities but does not replace humans.

Types of artificial intelligence : Artificial intelligence is normally categorized into four types. These categories are:

1. Reactive Machines,
2. Limited Memory,
3. Theory of Mind and
4. Self-awareness.

Arend Hintze proposed these four categories, an assistant profession of integrative biology and computer science engineering at Michigan State University.

Reactive Machines
Artificial intelligence that can make a prediction but has no memory and thus cannot use past experiences falls in the category of the reactive machine. The Deep Blue chess program developed by IBM is an example of a reactive device. In the 1990s, this program beat Garry Kasparov in chess. These types of artificial intelligence analyze inputs and choose the most strategic action. Normally, reactive machines designed for a task cannot be used in another different job. For example, AI designed for chess games cannot be used in card games.

Limited Memory
Artificial intelligence with the capability to use the experience to make future decisions falls in this category. The self-driving car is an excellent example of limited memory AI system. The AI of a self-driving vehicle observes the actions happening in the present and decides the next move. These observations are not stored permanently. They are erased continuously by new words.

Self-awareness
Self-awareness is a sense of self, having consciousness. The artificial intelligence system where machines can understand their current states (not just a sensor response) and use the previously stored information to infer others’ feelings fall in the self-awareness category. These types of AI are still under development.

Examples of AI technology
Artificial intelligence is incorporated into a variety of different types of technology. Here are the most well-known six technologies:

Automation
Automation is a technology that makes a system or process function automatically. For example, robotic process automation (RPA) can be programmed to perform repetitive and high-volume tasks. There is a difference between information technology (IT) automation and RPA. Normally IT automation remains at the software level, whereas RPA encompasses both software and hardware automation. Besides, RPA can adapt to changing environments.

Machine Learning
Machine learning is the science of getting a computer to perform without programming. We know computers are dumb and can serve only what it is programmed to function. Machine learning takes computers beyond this limitation. One of the most popular fields of machine learning is Deep Learning. We can say that Deep learning is the automation of predictive analytics. There are three types of machine learning algorithms. They are:

Supervised Learning:
A learning method where the correct or expected output is pre-defined in the dataset. The training data are labeled as ‘input’ and ‘correct output.’ The machines are trained to provide the work close to correct results during the training process.

Underserviced Learning
In this learning algorithm, the expected output is not defined in the dataset. In this approach, the machines are trained based on similarities and dissimilarities.

Reinforcement Learning
The training dataset is not labeled as input and correct output in reinforcement learning. However, after performing several actions, the AI system is given feedback. Based on the information, the system decides whether to keep existing characteristics or modify them.

Machine Vision

Machine vision is the science of enabling computers to see. This technology captures visual information using a digital camera. After that, the visible data is processed to generate meaningful data. For example, if a camera captures your image, it is a collection of pixel values located in the 2D spatial domain. If we use a face recognition algorithm to recognize the face on the image, it will be an application of machine vision. Sometimes, machine vision is compared to human visual perception (eyesight and visual processing in the brain). However, machine vision is not limited to biological boundaries. Signature identification, face recognition-based security systems, and medical image analysis are common machine vision applications. Commonly, machine vision and computer vision are used interchangeably. Computer vision refers to machine-based image processing. That means
computer vision is a subset of machine vision.

Natural Language Processing (NLP):
Natural language processing is the science of processing human language by computers. One of the most common examples of natural language processing is spam mail detection. This detection system looks at the email’s subject line and the text and can recognize spam and regular email. The current approaches to natural language processing are machine learning-based. Voice command, speech recognition, and automatic translation are a few examples of NLP applications.

Robotics
Robotics is a field of engineering that studies the design and manufacturing of robots. Robots are used in tasks that are not safe for humans. And they are perfect for performing repetitive duties. Robots are faster and can work accurately and consistently. Few of the common applications of robots are in assembly lines for car production, building mobile and computer motherboards, and space exploration.

Self-Driving Car
The self-driving car is an example of the application of artificial intelligence where computer vision, image recognition, and deep learning build the mechanical skill to pilot a vehicle. A self-driving car can travel to destinations without human intervention. They can also make quick and intelligent decisions in case of a sudden change in the environment. We cannot think of today’s world without artificial intelligence. It is everywhere. It has already gone beyond the capability of an individual to learn every branch of AI. That is why, nowadays, scientists and engineers work in different subbranches of AI. You need to understand hundreds of algorithms and thousands of technical terms to develop AI solutions. Not everyone wants to be an AI developer. But to understand today’s world, everyone must understand what artificial intelligence is. And this is exactly what you’ve learned from this article.

About Writer: Nuruzzaman Faruqui,Senior Lecturer, Department of Software Engineering, DIU