Information and Communication Technology · Classes 11-12 · Chapter 1 · 1.3

Contemporary Trends of ICT and 1.3.1 Artificial Intelligence.

You type a question and ChatGPT writes back. How does a machine manage that? In this note we open up artificial intelligence: what it is, the fields it works in, and how it learns.

1. Three steps of industrial revolution

Section 1.3 of the book opens with the industrial revolutions. The use of steam power began the first. The wide use of electric power was the second. And the use of electronics and ICT has started a new one, or several, as some put it.

The lesson of history is plain. Nations that took part in the earlier revolutions later led the world. So a nation will gain as much as it takes up this technology boldly.

2. What AI is, and why we need it

The idea of giving a machine human-like intelligence, and the ability to think or analyse, is called artificial intelligence.

Why do we need it? Simple computer programs are not enough for huge amounts of data. We need algorithms with which a computer finds a solution by thinking, and adapts to a changing situation as a human would. AI works with exactly these methods.

3. Five fields of AI

In the book's text there are five notable fields of AI: machine learning, robotics, computer vision, natural language processing (NLP) and speech processing. E-learning is not on the list.

Machine learning means training a computer so that it learns the rules of a system on its own, and learns from its mistakes to use that later. Robotics is the technology of making a robot or machine work automatically using AI.

NLP lets us exchange information with a computer in the languages people use, such as Bengali, English or Arabic. Computer vision is a way for a machine to process information from what its camera sees, as we do with our eyes. Speech processing is the technique of making the computer talk and listen. Now drop the five jobs into the five fields.

match them Recall the book's definition.

Each field's job

Pick up a job, then drop it in the field it belongs to.

Machine learning: learning rules itself. Robotics: automatic work. NLP: exchanging information in Bengali, English, Arabic. Computer vision: processing what a camera sees. Speech processing: talking and listening.

4. Neural networks and deep learning

The neural network is a popular AI method that works somewhat like the human brain. Mathematical artificial neurons, called perceptrons, are arranged in layers and connected, and that makes the network. Its job is to learn a function that gives a specific output from given inputs. It usually has three layers: the input layer, the hidden layer and the output layer. Use many hidden layers instead of one and even very complex functions can be learned. That is Deep Learning.

cut from the note Flip to see the meaning.

Layers and their jobs

TermInput layer

MeaningReceives the input.

TermHidden layer

MeaningIts values are tuned to give the right output.

TermOutput layer

MeaningGives the output of the system.

TermMany hidden layers

MeaningDeep learning, so complex functions can be learned.

AI systems are mainly built with the programming languages C/C++, Java, MATLAB, Python, SHRDLU, PROLOG, LISP, CLISP and R.

5. Three kinds of machine learning

Machine learning is considered the most successful field of AI. It is broadly divided into three kinds: supervised, unsupervised and reinforcement learning.

In supervised learning the machine is given many examples from which it learns what to do. For example, dog and cat pictures are shown and labelled as dogs and cats.

In unsupervised learning the computer is not told anything specific. It analyses a lot of data and finds the likenesses and differences itself. Without any names it will see that dogs and wolves are alike and differ from monkeys and chimpanzees.

In reinforcement learning nothing is taught separately and the computer works on its own. Afterwards it is only told how right or wrong it was, so next time it can change its behaviour or decision. That is why a chess computer sometimes makes a wrong move: it decides from information and experience, and learns from its mistakes to do better later.

Run the three kinds yourself: teaching with labels, grouping without labels, and teaching by right and wrong.

teach the machine pick a kind

Twelve animals, three ways to learn

The animals and their measurements are made up, to show the idea.

Now sort the examples into the three kinds.

sort them Think: labels or not, feedback or not.

Three kinds of machine learning

Pick up a chip, then drop it in a box. Ask: are names or labels given, and is the result told after the task?

Supervised: many examples, dog and cat pictures with names. Unsupervised: nothing is told, it sees the likeness of dogs and wolves itself. Reinforcement: works on its own, told right or wrong at the end.

6. Where it is put to work

Successful uses of AI include driverless cars and weather prediction. It is used in health, automobiles, finance, surveillance, education, gaming, agriculture, e-commerce, share trading, flying and running battlefields, among many other fields.

High-performance NLP software built with machine learning and deep learning includes ChatGPT, Microsoft Copilot and Google Bard. That answers our opening question: they handle language because they have learned it.

7. Check yourself

Three questions in the exam's style. Start with the first.

check yourself pick one answer

Question 1

Which technology expresses human thinking through machines?

Now the second.

check yourself pick one answer

Question 2

Where is artificial intelligence mainly used?

Take your time over the last one.

check yourself pick one answer

Question 3

Which is regarded as the most successful field in artificial intelligence?

The whole note on seven cards.

cut from the note tap to turn over

Seven cards

QuestionWhat is artificial intelligence?

AnswerThe idea of giving a machine human-like intelligence and the ability to think or analyse.

QuestionName four fields of AI.

AnswerMachine learning, robotics, computer vision, natural language processing (and speech processing).

QuestionWhat is machine learning?

AnswerTraining a computer so it learns rules on its own and learns from its mistakes.

QuestionWhich popular AI method works like the human brain?

AnswerThe neural network.

QuestionHow many layers does a neural network usually have, and which?

AnswerThree: the input layer, the hidden layer and the output layer.

QuestionWhat is using many hidden layers instead of one called?

AnswerDeep Learning.

QuestionIn which learning is the computer only told afterwards how right or wrong it was?

AnswerReinforcement learning.

Source: Information and Communication Technology, Classes 11-12 (NCTB), Chapter 1 “Information and Communication Technology: World and Bangladesh Perspective”, section 1.3.