Wikipedia AI Article Creation refers to the first encyclopedia article on artificial intelligence, published on the English Wikipedia in 2001. The article, titled "Artificial intelligence," provided a comprehensive overview of the field, defining AI as the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making. It described AI as a field of research in engineering, mathematics, and computer science that develops methods and software enabling machines to perceive their environment and use learning and intelligence to take actions that maximize chances of achieving defined goals.
The article outlined the traditional goals of AI research, including learning, reasoning, knowledge representation, planning, natural language processing, and perception, as well as support for robotics. It noted that AI researchers use techniques such as state space search, mathematical optimization, formal logic, artificial neural networks, and methods based on statistics, operations research, and economics. The article also acknowledged AI's interdisciplinary nature, drawing upon psychology, linguistics, philosophy, neuroscience, and other fields.
Historical Context
The article traced AI's origins to its founding as an academic discipline in 1956. It described the field's history of cycles of optimism followed by periods of disappointment and funding loss, known as AI winters. The article noted that funding and interest increased substantially after 2012, when graphics processing units (GPUs) began being used to accelerate neural networks, and deep learning outperformed previous AI techniques. This growth accelerated further after 2017 with the introduction of the transformer architecture. In the 2020s, an AI boom coincided with advances in generative AI, which became widespread and allowed for the creation and modification of media.
The article also discussed high-profile applications of AI, including advanced web search engines, chatbots, virtual assistants, autonomous vehicles, play and analysis in strategy games such as chess and Go, and content generation for images, audio, and videos. It mentioned that some companies, such as OpenAI, Google DeepMind, and Meta, aim to create artificial general intelligence (AGI) - AI that can complete nearly any cognitive task at least as well as a human.
Goals of AI Research
The article broke down the general problem of simulating intelligence into subproblems, each representing specific traits or capabilities that researchers expect an intelligent system to display. These included reasoning and problem-solving, knowledge representation, planning and decision-making, learning, natural language processing, perception, and support for robotics.
Reasoning and Problem-Solving
Early researchers developed algorithms that imitated step-by-step reasoning used by humans when solving puzzles or making logical deductions. By the late 1980s and 1990s, methods were developed for dealing with uncertain or incomplete information, employing concepts from probability and economics. However, many of these algorithms were insufficient for large reasoning problems due to a "combinatorial explosion," where they become exponentially slower as problems grow. The article noted that even humans rarely use step-by-step deduction; instead, they rely on fast, intuitive judgments. It also mentioned that reasoning models, a type of large language model trained to generate intermediate chains-of-thought, emerged in 2024 and allowed improved performance on complex problems in mathematics and coding, though they can produce incorrect outputs or "hallucinations."
Knowledge Representation
AI programs use knowledge to answer questions intelligently and make deductions about real-world facts. The article explained that formal knowledge representation and knowledge engineering use symbols to represent words, concepts, and things in the world. A knowledge base is a body of knowledge represented in a form usable by a program, and an ontology is the set of objects, relations, concepts, and properties used by a particular domain. Formal knowledge has been studied extensively since the 1970s, but the symbolic approach faces challenges such as the breadth of commonsense knowledge, the sub-symbolic form of most commonsense knowledge, and knowledge acquisition. The article noted that machine learning, particularly with large language models, does not require explicit symbolic knowledge and instead acquires knowledge from training on vast text corpora, solving some of these problems but struggling with accurate recall and valid reasoning.
Planning and Decision-Making
The article described an "agent" as any entity that perceives and takes actions in the world. A rational agent has goals or preferences and takes actions to achieve them. In automated planning, the agent has a specific goal; in automated decision-making, it has preferences and assigns a utility number to each situation. For each possible action, it calculates the expected utility, weighted by the probability of outcomes, and chooses the action with maximum expected utility. In classical planning, the agent knows the exact effects of actions, but in real-world problems, it may face uncertainty and must make probabilistic guesses and reassess. The article emphasized the importance of explanations for decisions to build trust, especially when decisions are relied upon. It also mentioned that preferences may be uncertain, and agents can learn them using methods like inverse reinforcement learning or seek information to improve them, with information value theory weighing the value of exploratory actions.
Techniques and Methods
To achieve these goals, AI researchers use a variety of techniques. The article highlighted state space search and mathematical optimization, formal logic, artificial neural networks, and methods based on statistics, operations research, and economics. It noted that since 2012, the use of GPUs to accelerate neural networks has been pivotal, and the transformer architecture introduced in 2017 further accelerated progress. The article also mentioned that modern AI can learn about a domain by running experiments, as when AlphaZero learns game strategy by playing against itself.
Ethical and Regulatory Considerations
The article addressed ethical concerns and potential harms from AI use, including AI safety, unintended consequences, long-term effects, environmental effects, and potential existential risks. It noted that these concerns have prompted discussions of AI regulation. The article also mentioned that AI's widespread adoption in the 2020s has led to debates about its societal impact, including issues of bias, privacy, and accountability.
Legacy and Impact
As an early encyclopedia entry, the article provided a foundational reference for understanding AI's scope and challenges. It has been continuously updated by the Wikipedia community to reflect advances in the field, such as the emergence of generative AI and deep learning. The article's structure, covering goals, techniques, and history, has influenced subsequent AI-related articles on Wikipedia and served as a resource for students, researchers, and the public. Its neutral, factual tone aligns with Wikipedia's standards, and its content has evolved to include recent developments like neural networks, large language models, and reinforcement learning (though the latter is not explicitly in the original 2001 text). The article remains a key entry point for readers seeking an overview of artificial intelligence.