Module 1 — The History & Landscape of AI

The story of artificial intelligence begins with Alan Turing, whose 1950 paper "Computing Machinery and Intelligence" posed the revolutionary question: "Can machines think?" The Turing Test proposed that a machine could be considered intelligent if a human interrogator couldn't distinguish its responses from a human's.

The field was officially born at the Dartmouth Conference in 1956, where John McCarthy coined the term "artificial intelligence." Early optimism produced programs that could prove theorems, play checkers, and understand simple language. Researchers predicted human-level AI within decades.

But reality proved harder. The first AI winter (1970s) came when early approaches hit fundamental limits. The second AI winter (late 1980s-90s) followed the collapse of expert systems hype. Each winter was triggered by overpromising and underdelivering.

The modern revolution began with three converging forces: big data (the internet generated massive training datasets), computational power (GPUs enabled training of deep neural networks), and algorithmic breakthroughs (deep learning techniques that could learn hierarchical representations). Key milestones: IBM Watson winning Jeopardy! (2011), AlphaGo defeating the world Go champion (2016), GPT-3 demonstrating language generation capabilities (2020), and ChatGPT making AI accessible to the general public (2022).

Understanding AI requires distinguishing three concepts:

Narrow AI (also called Weak AI or ANI) is designed to perform specific tasks. Every AI system that exists today is narrow AI — from Siri to self-driving cars to medical imaging analysis. A chess AI can beat any human at chess but cannot drive a car or write a poem. These systems are "intelligent" within their domain but have no general understanding.

Artificial General Intelligence (AGI) would match human cognitive abilities across all domains — reasoning, learning, creativity, social intelligence, and emotional understanding. AGI doesn't exist yet and remains one of the great unsolved challenges in computer science. Estimates for its arrival range from "within a decade" to "never," depending on who you ask.

Superintelligence (ASI) would surpass human intelligence in every domain. This concept, explored by philosopher Nick Bostrom, raises profound questions about control, alignment, and existential risk. If a system were smarter than humans in every way, could we ensure it pursues goals aligned with human values?

The current landscape is dominated by narrow AI that is becoming increasingly capable. Large Language Models (LLMs) like GPT-4 and Claude show surprising generality — they can write, reason, code, and analyse across many domains — but they still lack genuine understanding, common sense, and the ability to learn from experience the way humans do.

Today's AI excels at pattern recognition, prediction, content generation, and certain types of reasoning. Current capabilities include: natural language processing (translation, summarisation, question-answering), computer vision (image recognition, object detection, medical imaging), speech (transcription, synthesis, real-time translation), and decision-making (recommendation systems, fraud detection, autonomous driving).

Current limitations are equally important: AI systems can hallucinate (generate plausible but false information), lack common sense reasoning, cannot truly understand meaning (they process patterns in data), struggle with novel situations outside their training data, and lack ethical judgment.

The AI ecosystem includes major tech companies (Google DeepMind, OpenAI, Meta AI, Microsoft, Amazon, Apple), research institutions (Stanford HAI, MIT CSAIL, Berkeley AI Research), open-source community (Hugging Face, Meta's LLaMA, Mistral AI), and specialised startups across healthcare, finance, legal tech, and other verticals.

A notable trend is the democratisation of AI: tools that once required PhD-level expertise are now accessible through APIs and no-code platforms, enabling businesses of all sizes to leverage AI capabilities.