Artificial Intelligence

Artificial Intelligence was formally established as a field in 1955, when John McCarthy, a computer scientist at Stanford University, coined the term in his "Proposal for the Dartmouth Summer Research Project on Artificial Intelligence."

In this foundational text, McCarthy and his colleagues proposed the conjecture that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it".¹

While early definitions focused on the "science and engineering of making intelligent machines", modern academia often categorizes AI based on the specific goal of the system.

In their seminal textbook Artificial Intelligence: A Modern Approach, Russell and Norvig classify AI definitions into four distinct categories: systems that think like humans (cognitive modeling), systems that act like humans (Turing test approach), systems that think rationally (logic and laws of thought), and systems that act rationally (rational agents). Most contemporary research focuses on the latter—creating "rational agents" that act to maximize their expected success given available information.²

The field has evolved through several distinct paradigms. Early AI, often termed "Good Old-Fashioned AI" (GOFAI) or Symbolic AI, relied on explicit, human-programmed rules and logic to solve problems. However, this approach struggled with ambiguity and the complexity of the real world, leading to periods of reduced funding and interest known as "AI Winters."

In recent decades, the field has shifted toward Connectionism and Machine Learning, where systems learn patterns from data rather than following pre-programmed rules. This shift was driven by the resurgence of artificial neural networks—computational models inspired by the biological neural networks of the human brain.

The modern explosion of AI capabilities is largely attributed to Deep Learning, a subfield of machine learning that utilizes multi-layered neural networks to perform complex feature extraction. This paradigm shift was comprehensively documented by LeCun, Bengio, and Hinton in their landmark 2015 review in Nature, which established deep learning as the foundation for contemporary AI advances.^{3 4}

Today, AI encompasses a vast spectrum of technologies that have surpassed human performance in specific, narrow domains (Artificial Narrow Intelligence, or ANI).

Computer Vision: In healthcare, AI models have demonstrated the ability to outperform human specialists in diagnostic tasks. Deep learning has achieved significant breakthroughs in medical imaging: Esteva et al. (2017) showed that convolutional neural networks could achieve dermatologist-level classification of skin cancer from dermoscopic images, published in Nature. Similarly, McKinney et al. (2020) published a landmark study in Nature demonstrating that an AI system achieved higher sensitivity and specificity than radiologists in breast cancer screening across multiple international datasets, reducing both false positives and false negatives.^{5 6}

Natural Language Processing (NLP): The development of the Transformer architecture revolutionized machine translation and language understanding. Vaswani et al.'s seminal 2017 paper "Attention Is All You Need," presented at the Neural Information Processing Systems (NeurIPS) conference, introduced the Transformer model which has become foundational to modern natural language processing systems, allowing models to process text with near-human fluency and context awareness.⁷

Despite these advancements, current systems remain distinct from Artificial General Intelligence (AGI)—hypothetical systems that would possess the ability to understand, learn, and apply knowledge across a wide variety of tasks at a human level.