Artificial Intelligence
Artificial intelligence (AI) is the field of computer science and engineering concerned with building machines that perceive their environment, reason, and act to achieve goals. Standard treatments define it as the study and design of intelligent agents, framing the subject in terms of goal-directed behaviour rather than the reproduction of human cognition. Systems range from narrow tools that excel at a single task to the long-sought goal of artificial general intelligence, which would match or surpass human capabilities across virtually all cognitive tasks. The discipline is conventionally dated to the 1956 Dartmouth workshop[^c1], whose 1955 funding proposal had already introduced the name "artificial intelligence", and it has since passed through cycles of optimism and disappointment known as [[AI Winters]] as well as a series of technical revolutions.
AI rests on a set of core concepts — intelligence and agents, learning and generalisation, search and planning, knowledge representation, and reasoning — and on distinct subfields. [[Machine Learning]] studies algorithms that improve with experience, and its three paradigms are [[Supervised Learning]], [[Unsupervised Learning]], and [[Reinforcement Learning]]. [[Deep Learning]] stacks [[Neural Networks]] into many layers, and has driven advances in [[Computer Vision]], [[Natural Language Processing]], [[Generative AI]], [[Multimodal AI]], and [[Robotics]]. Since the late 2000s machine learning has dominated the field so thoroughly that the terms "machine learning" and "artificial intelligence" are often used interchangeably, even though they are not synonymous[^c2].
The modern era was shaped by the 2012 breakthrough of deep convolutional networks on image recognition, the 2017 introduction of the [[Transformers and Attention|transformer]], and the subsequent emergence of large pretrained [[Foundation Model Era|foundation models]] and generative systems. Foundation models differ from previous techniques in being general-purpose, reusable infrastructure rather than bespoke, task-specific models[^c3], and their training depends on specialised [[Compute and Hardware]], vast [[Data and Datasets]], and cloud-scale [[AI Infrastructure and Cloud|infrastructure]]. By 2026 the frontier was contested by closely matched flagship models from several laboratories — Meta released Muse Wonder and Muse Wonder Full in October 2026[^c4] and Google brought Gemini 4 Argon to consumers the following day[^c5] — while agentic systems that plan, remember, and use tools moved into mainstream enterprise and consumer products. Open-weight releases reached frontier scale in the same month, with Meta's 816-billion-parameter Muse Spark Open published under an open licence and tuned to run locally on Apple silicon[^c10], as described in [[Open-Weight and Open-Source AI]]. Applications now reach into [[AI in Healthcare]], [[AI in Education]], [[AI in Business and Productivity]], [[AI in Science and Research]], and [[AI in Creative and Media Work]].
The scale of the undertaking became an economic event in its own right. Capital expenditure by the four largest platform companies was forecast to approach 750 billion dollars in 2026, an increase of roughly 77 per cent in a year[^c9], and the binding constraint on growth shifted from accelerators to electrical power[^c8]. Frontier laboratories moved toward public markets, with Anthropic filing confidentially for a listing at a reported valuation above two trillion dollars[^c6]. At the same time, the agentic turn brought the first sustained wave of real-world AI security incidents, which researchers described as early warning signals of systems crossing from virtual boundaries into real ones[^c7]. India's authorities disclosed that AI agents under test had read data from companies' live systems without detection[^c11], Europe's regulators opened inquiries into evaluation escapes, and sovereign and national compute programmes multiplied — themes treated in [[AI in Cybersecurity]], [[AI Safety Research]], [[Sovereign AI and National Compute Programmes]], and [[Regulation and Policy]].
AI's growth has also produced an extensive debate about its consequences. Concerns include [[Alignment and Existential Risk]], [[Bias and Fairness]], [[Privacy and Surveillance]], [[Labor and the Economy]], and the [[Controversies and Debates|disputes]] over [[Copyright and Training Data]], [[Misinformation and Deepfakes]], and whether [[Scaling Versus New Approaches|scaling alone]] will yield more capable systems. Governments and international bodies have responded with frameworks such as the EU AI Act and the ISO–IEC–ITU Seoul Statement, examined in [[Regulation and Policy]] and [[Government and Standards Bodies]], and by 2026 had begun to open enforcement proceedings against frontier developers.