Coursify

Introduction to Artificial Intelligence: Principles, History, and Classification

Introduction to Artificial Intelligence: Principles, History, and Classification

Verified Sources
Jun 12, 2026

Artificial Intelligence (AI) represents one of the most profound technological paradigm shifts in human history . At its core, AI is a branch of computer science dedicated to building systems capable of performing tasks that traditionally require human cognitive faculties—such as learning, reasoning, problem-solving, perception, and linguistic understanding .

Rather than relying on static, hard-coded instructions, modern AI systems utilize algorithms to analyze vast datasets, recognize complex patterns, and make autonomous decisions.

Core Paradigms: Symbolic AI vs. Connectionist AI

Historically, AI has been divided into two primary paradigms:

  1. Symbolic (Good Old-Fashioned AI - GOFAI): Relies on explicit, logic-based representations of knowledge and rule engines. It excels at structured, deterministic tasks (like chess or mathematical proofs) but struggles with noisy, real-world data.
  2. Connectionist (Machine Learning & Neural Networks): Inspired by biological brains, this approach uses mathematical networks to learn representations directly from raw data.

To formalize how an AI agent operates, we often model it as a rational entity interacting within an environment to maximize its expected utility . This interaction loop can be represented mathematically and visually:

max⁡πE[∑t=0∞γtRt]\max_{\pi} \mathbb{E} \left[ \sum_{t=0}^{\infty} \gamma^t R_t \right]

Where π\pi represents the decision policy, γ\gamma is the discount factor (0≤γ<10 \le \gamma < 1), and RtR_t is the reward at time step tt.

Footnotes

  1. Artificial intelligence (AI) | Definition, Examples, Types, Applications - Britannica's comprehensive overview of the definitions, historical roots, and core components of AI. ↩ ↩2 ↩3

AI, Machine Learning, Deep Learning and Generative AI Explained

The Historical Evolution of AI

The Birth of Artificial Neurons

1943

Warren McCulloch and Walter Pitts publish 'A Logical Calculus of Ideas Immanent in Nervous Activity,' introducing the first mathematical model of an artificial neural network ."

Footnotes

  1. Artificial intelligence: definition, history and applications - Theodo's summary of the McCulloch-Pitts neural model and the founding of AI at Dartmouth. ↩

The Turing Test Proposed

1950

Alan Turing publishes 'Computing Machinery and Intelligence,' proposing the [Turing Test]{def='A test developed by Alan Turing to evaluate a machine's ability to exhibit intelligent behavior equivalent to a human.'} as a benchmark for machine intelligence ."

Footnotes

  1. What is the history of artificial intelligence (AI)? - Tableau's detailed analysis of AI's historical milestones, including McCarthy's LISP and the AI Winters. ↩

The Dartmouth Workshop

1956

John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon organize the Dartmouth Summer Research Project, where the term 'Artificial Intelligence' is officially coined , ."

Footnotes

  1. What is the history of artificial intelligence (AI)? - Tableau's detailed analysis of AI's historical milestones, including McCarthy's LISP and the AI Winters. ↩

  2. Artificial intelligence: definition, history and applications - Theodo's summary of the McCulloch-Pitts neural model and the founding of AI at Dartmouth. ↩

The First AI Winter

1974 - 1980

Overpromised capabilities lead to public disappointment. Funding from agencies like DARPA is severely cut, leading to a period of stagnation known as the 'AI Winter' ."

Footnotes

  1. What is the history of artificial intelligence (AI)? - Tableau's detailed analysis of AI's historical milestones, including McCarthy's LISP and the AI Winters. ↩

Deep Blue Defeats Kasparov

1997

IBM's Deep Blue supercomputer defeats reigning World Chess Champion Garry Kasparov, proving the viability of heuristic-search-based symbolic AI ."

Footnotes

  1. What is the history of artificial intelligence (AI)? - Tableau's detailed analysis of AI's historical milestones, including McCarthy's LISP and the AI Winters. ↩

The Deep Learning Boom

2012

AlexNet, a deep convolutional neural network designed by Alex Krizhevsky, wins the ImageNet competition by a massive margin, initiating the modern Deep Learning revolution."

Generative AI Era

2020s

Large Language Models (LLMs) achieve mainstream global adoption, transforming creative, analytical, and computational industries ."

Footnotes

  1. Artificial Intelligence Market to Grow at 36.6% CAGR - Grand View Research report highlighting global AI market sizes, growth projections, and economic impacts through 2030. ↩

Artificial Narrow Intelligence (ANI), also known as 'Weak AI,' is the only type of AI that currently exists in the real world . These systems excel at their designated domain—such as facial recognition, language translation, or playing Go—but cannot apply their cognitive abilities to any other task. If an ANI chess program is asked to predict the weather, it cannot function, as its mathematical architecture is strictly bounded by its training domain .

Footnotes

  1. 3 Types of Artificial Intelligence - ANI , AGI and ASI - Kaggle's breakdown of the differences between Narrow, General, and Super Artificial Intelligence. ↩ ↩2

Projected Global AI Market Growth (2024 - 2030)

Market size in billions of USD ($) showing a compound annual growth rate (CAGR) of 36.6% .

Footnotes

  1. Artificial Intelligence Market to Grow at 36.6% CAGR - Grand View Research report highlighting global AI market sizes, growth projections, and economic impacts through 2030. ↩

The Machine Learning Lifecycle

  1. 1
    Step 1

    Raw data is collected from diverse sources and cleaned. This step involves handling missing values, removing outliers, and normalizing features so that the input vectors XX are mathematically consistent.

  2. 2
    Step 2

    Domain experts and data scientists extract the most informative mathematical attributes (features) from the raw data, reducing dimensionality and highlighting patterns that make learning easier.

  3. 3
    Step 3

    An appropriate algorithm (such as a Support Vector Machine or a Deep Neural Network) is chosen. The model parameters—represented as weights W\mathbf{W} and biases b\mathbf{b}—are initialized, often using random distributions.

  4. 4
    Step 4

    The training data is fed through the model. The network performs matrix multiplications and applies non-linear activation functions to compute a prediction: Y^=σ(W⋅X+b)\hat{Y} = \sigma(\mathbf{W} \cdot X + \mathbf{b}).

  5. 5
    Step 5

    The model's prediction Y^\hat{Y} is compared against the ground-truth label YY using a loss function L(Y,Y^)L(Y, \hat{Y}), which quantifies the model's error.

  6. 6
    Step 6

    The error is sent backward through the network using the chain rule of calculus to compute the gradient of the loss with respect to each weight: ∂L∂W\frac{\partial L}{\partial \mathbf{W}}. An optimization algorithm, such as Gradient Descent, updates the weights to minimize the loss: W←W−α∂L∂W\mathbf{W} \leftarrow \mathbf{W} - \alpha \frac{\partial L}{\partial \mathbf{W}}, where α\alpha is the learning rate.

The AI Venn Diagram

Keep this nested hierarchy in mind: Artificial Intelligence is the broad, overarching scientific field. Machine Learning (ML) is a subfield of AI focused on statistical learning from data. Deep Learning (DL) is a further subfield of ML that specifically utilizes multi-layered neural networks to automatically extract features from raw data.

The Hallucination Vector

Because modern Large Language Models (LLMs) operate on conditional probability distributions—predicting the next most likely token P(Tn∣T1,…,Tn−1)P(T_{n} | T_{1}, \dots, T_{n-1})—they lack a built-in mechanism for verifying factual truth. This can result in hallucinations, where the model generates highly confident but completely fabricated assertions.

Ethical and Practical Challenges in AI

Knowledge Check

Question 1 of 3
Q1Single choice

Which of the following best describes the fundamental difference between Symbolic AI and Connectionist AI?

Explore Related Topics

1

AI Roadmap 2026: From Foundations to Frontier

The AI Roadmap 2026 maps the shift from standalone large language models to interconnected, agentic and multimodal AI ecosystems, outlining key trends, the modern AI stack, essential skills, and a 12‑month learning pathway to become job‑ready.

  • Five macro trends: agentic AI, multimodal AI, AI‑bubble deflation, governance‑as‑code, and AI economic dashboards.
  • Six‑layer stack: reasoning LLMs, RAG & vector DBs, agent frameworks (LangChain, MCP), guardrails, memory/state, and evaluation/observability.
  • In‑demand transversal skills: Python/ML frameworks, LLM/GenAI, cloud & MLOps, agent development, RAG/vector databases, and AI governance/ethics.
  • Defined career tracks (AI Engineer, ML Engineer, Deep Learning Engineer, Research Engineer) with salary ranges and role‑specific tech stacks.
  • Career value grows multiplicatively: Career Value=∑i=1n(Skill Depthi×Market Demandi)\text{Career Value} = \sum_{i=1}^{n} (\text{Skill Depth}_i \times \text{Market Demand}_i).
2

AI vs Human Teachers: A Comprehensive Analysis

The module examines AI versus human teachers, advocating a hybrid approach where AI automates routine, personalized tasks while teachers supply emotional, mentorship, and critical‑thinking support.

  • AI provides 24/7 availability, adaptive personalization, instant objective feedback, and scalability, freeing ~10 hrs/week of teacher workload.
  • Human teachers contribute empathy, mentorship, cultural interpretation, ethical judgment, and social modeling—capabilities AI cannot replicate.
  • Studies show AI use raises engagement (β=0.48\beta = 0.48, p<0.001p < 0.001) but excessive reliance harms critical‑thinking skills.
  • Optimal effectiveness combines AI efficiency with human depth: Educational Effectiveness=f(AI Efficiency)+g(Human Depth)\text{Educational Effectiveness}=f(\text{AI Efficiency})+g(\text{Human Depth}).
3

Generative AI

Generative AI comprises models that learn data distributions to create new text, images, audio, code, or video, driven mainly by transformer‑based language models and diffusion image models.

  • Core architectures are large‑scale transformers for language and diffusion models for image synthesis, approximating p(x)p(x), p(x∣c)p(x\mid c) or p(y∣x)p(y\mid x).
  • Foundation models are pretrained on massive corpora and adapted via prompting, fine‑tuning, or retrieval‑augmented generation for diverse downstream tasks.
  • 2023 saw 25.225.2 billion private investment, 149 new foundation models, and a rise to 65.7%65.7\% open‑source releases, though frontier models remain costly.
  • Major risks include confabulation, bias, privacy leakage, copyright disputes, and deepfake misuse, requiring systematic governance.
  • Responsible deployment combines data provenance, RAG grounding, safety layers, human oversight, and continuous monitoring.
Browse all research articles