Artificial Intelligence, Machine Learning and Deep Learning

1. What is Artificial Intelligence?

Artificial Intelligence is a branch of computer science that focuses on creating systems capable of performing tasks that normally require human intelligence. These tasks include understanding language, recognising images, solving problems, making decisions, and learning from experience.

AI does not mean that every machine thinks exactly like a human. Most modern AI systems are designed for a specific purpose, such as recommending videos, detecting spam emails, translating languages, or helping a customer through a chatbot.

Examples of AI in daily life

• Google Maps suggesting a faster route based on traffic.
• Email services filtering unwanted messages into spam.
• Voice assistants answering questions or setting reminders.
• Streaming platforms recommending films or songs.
• Banking systems detecting potentially fraudulent transactions.

Brief history of AI

The term Artificial Intelligence was coined by John McCarthy in connection with the Dartmouth Summer Research Project on Artificial Intelligence in 1956. The workshop is widely regarded as an important founding event for AI as an academic field.

AI has many important contributors, including John McCarthy, Marvin Minsky, Claude Shannon, Allen Newell, Herbert Simon, Arthur Samuel, and others. Therefore, it is more accurate to describe John McCarthy as a key pioneer of AI rather than the only “father” of the field.

2. Major Areas of Artificial Intelligence

AI is a broad subject made up of several related areas. Some of the major areas are:

1. Machine Learning: Learning patterns from data.
2. Deep Learning: Using multi-layer neural networks to learn complex patterns.
3. Natural Language Processing: Helping computers understand and generate human language.
4. Computer Vision: Enabling computers to understand images and videos.
5. Robotics: Designing machines that can sense, decide, and act in the physical world.
6. Expert Systems: Programs that use rules and knowledge to solve specialised problems.
7. Fuzzy Logic: Reasoning with degrees of truth instead of only true or false values.

3. Common Types of Artificial Intelligence

A commonly taught, functionality-based classification divides AI into four types. This is a useful learning framework, although it is not the only way to classify AI.

1. Reactive Machines: These systems respond only to current inputs and do not store past experiences. A basic chess-playing system is often used as an example.
2. Limited Memory AI: These systems use recent or historical data to make decisions. Many current machine-learning systems, including recommendation and driving-assistance systems, fit this category.
3. Theory of Mind AI: A theoretical form of AI that would understand emotions, beliefs, and intentions. Fully developed systems of this type do not currently exist.
4. Self-Aware AI: A hypothetical form of AI that would possess awareness of itself. This remains a concept rather than a current technology.

4. What is Machine Learning?

Machine Learning is a branch of AI in which algorithms learn patterns from data. Instead of writing a separate rule for every possible situation, developers provide examples or feedback so that a model can improve its predictions or decisions.

For example, an email-spam model can learn from thousands of labelled emails. Over time, it identifies patterns that are common in unwanted messages and uses those patterns to classify new emails.

The term Machine Learning was coined by Arthur Samuel in 1959. He is known for developing an early self-learning checkers program.

Machine Learning and Data Mining

Machine Learning and Data Mining are related but not identical. Machine Learning focuses on building models that learn from data, while Data Mining focuses on discovering useful patterns, relationships, or trends in large datasets. Machine-learning methods are often used during data-mining work.

5. Types of Machine Learning

The three most commonly discussed types of Machine Learning are supervised learning, unsupervised learning, and reinforcement learning.

Supervised Learning

Supervised learning uses labelled data. Each training example includes an input and the correct output. The model learns the relationship between them.

Example: Training a model using past house prices to predict the price of a new house.

Unsupervised Learning

Unsupervised learning uses data without labelled answers. The algorithm looks for natural patterns, groups, or unusual records in the data.

Example: Grouping customers into segments based on their purchasing behaviour.

Reinforcement Learning

Reinforcement learning is based on trial and error. An agent takes actions in an environment and receives rewards or penalties. Over time, it learns which actions are most useful for achieving a goal.

Example: Training a game-playing program, a robot-navigation system, or a traffic-signal controller.

6. What is Deep Learning?

Deep Learning is a specialised subset of Machine Learning that uses neural networks with multiple layers. These layers can learn complex patterns from large amounts of data.

Deep-learning systems are especially useful for tasks involving images, speech, video, and natural language. For example, a deep-learning model can learn to identify objects in photographs or convert spoken words into text.

Basic neural-network layers

• Input layer: Receives the original data, such as pixels, audio signals, or numbers.
• Hidden layers: Process information and learn useful patterns from the data.
• Output layer: Produces the final prediction, classification, or result.

Applications of Deep Learning

• Image and facial recognition
• Speech recognition and voice assistants
• Medical-image analysis
• Fraud detection
• Language translation
• Driver-assistance and autonomous-vehicle research

7. AI vs Machine Learning vs Deep Learning

8. Programming Languages Used in AI

Different languages are used in AI depending on the project, performance needs, available libraries, and development environment.

• Python: Widely used because of its readable syntax and libraries such as NumPy, pandas, scikit-learn, PyTorch, and TensorFlow.
• R: Common in statistics, data analysis, and research.
• Java and C++: Often used in large applications where performance and scalability are important.
• JavaScript: Useful for AI features that run in web applications.
• Lisp: Historically important in AI research. It was created by John McCarthy in 1958.
• Prolog: Known for logic programming and symbolic reasoning.
• Julia: Used in numerical computing and scientific applications.

Important note: IPL and Lisp are different languages. IPL was an early language used in AI research, while Lisp was later created by John McCarthy.

9. AI, Robotics and Chatbots

Robotics

Robotics combines mechanical engineering, electronics, programming, sensors, and AI. A robot may use AI to recognise objects, plan a route, respond to voice commands, or make decisions from sensor data.

A robot arm is commonly called a manipulator. Its actuators create movement, while its end effector is the part that interacts with an object, such as a gripper or welding tool.

Unimate began work at a General Motors plant in 1961 and is widely recognised as the first industrial robot.

Chatbots and virtual assistants

Chatbots are programs designed to communicate with users through text or voice. Modern chatbots often use Natural Language Processing and, in some cases, large language models.

10. Advantages and Limitations of AI

Advantages

• Can process large amounts of data quickly.
• Can automate repetitive tasks.
• Can improve consistency in well-defined tasks.
• Can assist experts in fields such as healthcare, finance, education, and engineering.
• Can operate continuously without fatigue.

Limitations

• AI systems can produce incorrect or biased results if the data or design is poor.
• Many systems require significant data, computing power, and careful testing.
• AI does not automatically understand context, emotions, or ethics like a human being.
• Privacy and security must be considered when personal data is used.
• Important decisions should include appropriate human oversight.

11. Key Takeaways

• AI is the broad field of making computers perform intelligent tasks.
• Machine Learning is a subset of AI that learns patterns from data.
• Deep Learning is a subset of ML that uses multi-layer neural networks.
• Supervised, unsupervised, and reinforcement learning are major ML approaches.
• AI is used in healthcare, education, finance, entertainment, cybersecurity, robotics, and many everyday applications.

12. Frequently Asked Questions

Is AI the same as Machine Learning?

No. AI is the broader field, while Machine Learning is one method used to build AI systems.

Is Deep Learning different from Machine Learning?

Deep Learning is a specialised part of Machine Learning that uses neural networks with multiple layers.

What are the three major types of Machine Learning?

Supervised learning, unsupervised learning, and reinforcement learning.

Who coined the term Artificial Intelligence?

John McCarthy coined the term in connection with the Dartmouth AI research project in 1956.

What is an example of supervised learning?

A system trained on labelled emails to classify new emails as spam or not spam is an example of supervised learning.