Artificial Intelligence (AI): Introduction & Types
Evelyn Clarke
โก Smart Summary
Artificial Intelligence enables machines to perform cognitive functions such as perceiving, learning, reasoning, and problem-solving like humans. This tutorial covers the definition, history, goals, subfields, and types of AI, and explains how it differs from machine learning and why it is booming today.
What is Artificial Intelligence (AI)?
AI (Artificial Intelligence) is a machine’s ability to perform cognitive functions the way humans do, such as perceiving, learning, reasoning, and solving problems. The benchmark for AI is the human level in terms of reasoning, speech, and vision.
In this Artificial Intelligence tutorial, you will learn the AI basics covered in the sections below.
How AI Creates Business Value
Today, AI is used across almost all industries, giving a technological edge to companies that integrate it at scale. According to McKinsey, AI has the potential to create $600 billion of value in retail and bring 50% more incremental value in banking compared with other analytics techniques. In transport and logistics, the potential revenue jump is up to 89%.
For example, if an organization uses AI for its marketing team, it can automate mundane, repetitive tasks so sales representatives can focus on relationship building and lead nurturing. A company called Gong provides a conversation-intelligence service: each time a sales representative makes a call, the machine records, transcribes, and analyzes the conversation, and managers use AI analytics and recommendations to shape a winning strategy.
In short, AI provides cutting-edge technology to handle complex data that a human cannot process alone. It automates redundant jobs so workers focus on high-value tasks, and when implemented at scale it reduces costs and increases revenue.
History of Artificial Intelligence
Artificial Intelligence is a buzzword today, but the term is not new. In 1956, experts from different backgrounds organized a summer research project on AI. Four bright minds led it: John McCarthy (Dartmouth College), Marvin Minsky (Harvard University), Nathaniel Rochester (IBM), and Claude Shannon (Bell Telephone Laboratories). Here is a brief history of AI.
| Year | Milestone / Innovation |
|---|---|
| 1923 | Karel ฤapek’s play “Rossum’s Universal Robots” introduced the word “robot” into English. |
| 1943 | Foundations for neural networks were laid. |
| 1945 | Isaac Asimov, a Columbia University alumnus, used the term “robotics”. |
| 1956 | John McCarthy first used the term “Artificial Intelligence”, and the first running AI program was demonstrated. |
| 1964 | Danny Bobrow’s MIT dissertation showed how computers could understand natural language. |
| 1969 | Stanford Research Institute developed Shakey, a robot with locomotion and problem-solving. |
| 1979 | The Stanford Cart, the world’s first computer-controlled autonomous vehicle, was built. |
| 1990 | Significant demonstrations in machine learning. |
| 1997 | The Deep Blue chess program beat world champion Garry Kasparov. |
| 2000 | Interactive robot pets became commercially available; MIT displayed Kismet, a robot that expresses emotions. |
| 2006 | AI entered the business world; companies like Facebook, Netflix, and Twitter began using it. |
| 2012 | Google launched “Google Now”, an Android feature offering predictions. |
| 2018 | IBM’s “Project Debater” debated complex topics with master debaters and performed exceptionally well. |
Goals of Artificial Intelligence
The main goals of AI are:
- Reduce the time needed to perform specific tasks.
- Make it easier for humans to interact with machines.
- Facilitate more natural and efficient human-computer interaction.
- Improve the accuracy and speed of medical diagnoses.
- Help people learn new information more quickly.
- Enhance communication between humans and machines.
Subfields of Artificial Intelligence
Here are the important subfields of Artificial Intelligence.
Machine Learning: The art of studying algorithms that learn from examples and experience. It is based on identifying patterns in data and using them for future predictions. Unlike hard-coded rules, the machine learns the rules itself.
Deep Learning: A sub-field of machine learning that uses multiple layers to learn from data. The depth of the model is the number of layers it contains; for example, Google’s LeNet model for image recognition has 22 layers.
Natural Language Processing (NLP): The branch of AI that enables machines to read, understand, and generate human language. It powers applications such as chatbots, translation, sentiment analysis, and speech recognition.
Expert Systems: An interactive, reliable computer-based decision-making system that uses facts and heuristics to solve complex problems within a specific domain. It aims to solve the most difficult issues in that domain.
Fuzzy Logic: A many-valued logic in which truth values can be any real number between 0 and 1. It handles the concept of partial truth for situations where a statement is neither fully true nor fully false.
Types of Artificial Intelligence
AI can be classified two ways โ by capability and by technique.
By capability:
- Narrow AI: Performs a single dedicated task intelligently, such as image tagging or voice assistants.
- General AI: Can perform any intellectual task as efficiently as a human (still theoretical).
- Super AI: A hypothetical level where machines surpass human intelligence.
By technique:
- Rule-based AI: Applies a set of pre-determined rules to an input data set to produce a corresponding output.
- Decision tree AI: Similar to rule-based AI but allows branching and looping to weigh different options.
- Neural networks: Layered models inspired by the human brain that learn complex patterns from data, and underpin robotics with reasoning, planning, and learning abilities.
AI vs Machine Learning
AI and machine learning are often used interchangeably, but they are not the same. AI is the science of training machines to perform human tasks; the term was coined in the 1950s. Machine learning is a subset of AI in which a machine learns patterns from data rather than being explicitly programmed.
| Aspect | Artificial Intelligence | Machine Learning |
|---|---|---|
| Scope | Broad science of mimicking human intelligence | A subset of AI focused on learning from data |
| Goal | Simulate human reasoning and decision-making | Find patterns and make predictions |
| Approach | Rules, logic, and learning combined | Data-driven training with examples |
| Example | A self-driving car system | The image-recognition model inside it |
You can also read the difference between Deep Learning, Machine Learning, and AI.
Where is AI Used? Examples
AI has broad applications across industries.
- Eliminating repetitive tasks: AI can repeat a task continuously without fatigue and is indifferent to the work it carries out.
- Improving existing products: Firms add AI to enhance product functionality rather than build from scratch. For example, Facebook once required manual photo tagging; today AI suggests friends to tag automatically.
AI is used in every industry โ from marketing and supply chain to finance and food processing. According to a McKinsey survey, financial services and high-tech communication lead AI adoption.
Why is AI Booming Now?
Neural networks have existed since the 1990s, following the seminal paper by Yann LeCun, but they became famous around 2012. Three critical factors drive today’s AI boom.
Hardware
Over the last two decades, CPU power has exploded, letting users train small deep-learning models on a laptop. However, computer vision and large deep-learning models need more power. Thanks to investment from NVIDIA and AMD, a new generation of GPUs (graphical processing units) allows parallel computation, spreading work across several GPUs to speed up calculations. For instance, an NVIDIA TITAN X can train an ImageNet model in two days versus weeks on a traditional CPU, and big companies use GPU clusters such as the NVIDIA Tesla K80 to cut data-center costs and improve performance.
Data
Deep learning is the structure of the model, and data is the fluid that brings it alive. Without data, nothing can be done. Modern storage technology makes it easy to keep huge amounts of data in a data center, and the internet makes data collection and distribution available to feed machine-learning algorithms.
Apps like Flickr and Instagram hold millions of tagged pictures that can train a neural network to recognize objects without manual labeling. AI combined with data is the new gold: the company with the most data has a competitive advantage. The world creates about 2.2 exabytes (2.2 billion gigabytes) of data every day, and diverse data sources help models find patterns and learn at scale.
Algorithm
Hardware is more powerful than ever and data is easily accessible, but more accurate algorithms make neural networks reliable. Early neural networks were simple matrix multiplications without deep statistical properties; since 2010, remarkable discoveries have improved them. AI uses progressive learning algorithms that let the data do the programming, so a computer can teach itself to perform tasks such as detecting anomalies or acting as a chatbot.
