{"licence":{"name":"CC BY-SA 4.0","spdx":"CC-BY-SA-4.0","url":"https://creativecommons.org/licenses/by-sa/4.0/","attribution":"Atlas, a bilingual technical dictionary (https://cmaintz.github.io/tech-atlas/)"},"id":"ai/artificial-intelligence","url":{"en":"https://cmaintz.github.io/tech-atlas/en/terms/ai/artificial-intelligence/","da":"https://cmaintz.github.io/tech-atlas/da/terms/ai/artificial-intelligence/"},"term":{"en":"Artificial intelligence (AI)","da":"Kunstig intelligens (AI)"},"aka":{"en":["AI"],"da":["AI"]},"domain":["ai"],"cluster":"ml-fundamentals","layer":"theory","status":"current","era":1956,"summary":{"en":"Computer systems that do tasks we normally link to human thinking, such as spotting patterns, answering questions or making choices.","da":"Computersystemer, der løser opgaver, vi forbinder med menneskelig tænkning, fx at genkende mønstre, svare på spørgsmål eller træffe valg."},"body":{"formal":{"en":"A broad field and a label for systems that, for a set of goals chosen by people, produce outputs such as predictions, content, advice or decisions that affect the world around them, with some level of independence.","da":"Et bredt fagområde og en betegnelse for systemer, der ud fra mål sat af mennesker frembringer resultater som forudsigelser, indhold, anbefalinger eller beslutninger, der påvirker omgivelserne, med en vis grad af selvstændighed."},"plain":{"en":"An umbrella word, like “vehicle”, covering everything from a simple filter that sorts junk mail by fixed rules to a chat assistant that writes whole reports.","da":"Et paraplyord ligesom “køretøj” - det dækker alt fra et simpelt filter, der sorterer uønsket post efter faste regler, til en chatassistent, der skriver hele rapporter."},"inPractice":{"en":"A supplier tells a region's purchasing team that its new HR tool “uses AI”; the team asks whether it actually sorts job applications, writes text or makes decisions about staff.","da":"En leverandør fortæller en regions indkøbsafdeling, at dens nye HR-værktøj “bruger AI”; afdelingen spørger, hvad det faktisk gør - sorterer ansøgninger, skriver tekst eller træffer beslutninger om medarbejdere."},"whyItMatters":{"en":"The label alone says little about risk; laws such as the EU AI Act judge a system by what it is used for, so buyers and users must ask what it really does.","da":"Ordet alene siger ikke meget om risikoen; love som EU's AI-forordning vurderer et system efter, hvad det bruges til, så købere og brugere må spørge, hvad det reelt gør."}},"deepDive":{"en":"The term was coined by John McCarthy in the 1955 proposal for the 1956 Dartmouth Summer Research Project, which conjectured that every aspect of learning or intelligence could in principle be described precisely enough for a machine to simulate it. Alan Turing had already framed the question operationally in 1950 with the imitation game. From the start the field split into two traditions: symbolic AI (logic, search, knowledge representation, planning), and connectionist AI based on networks of simple units trained from data. Symbolic methods dominated until the 1980s, peaking with rule-based expert systems; their brittleness and maintenance cost, together with over-promising, contributed to two funding contractions usually called the AI winters (mid-1970s and late 1980s).\n\nRussell and Norvig define the field around rational agents: systems that perceive an environment and act to maximise an expected performance measure. This framing covers search algorithms, constraint solvers, probabilistic reasoning such as Bayesian networks, planning, robotics and machine learning under one roof. Since the 2010s, machine learning and in particular deep learning has become the dominant technique, which is why everyday usage now treats AI and ML as near-synonyms, and more recently AI and large language models. The two are not interchangeable: a route planner using A* search or a tax-rules engine is AI in the textbook sense but involves no learning.\n\nLegal definitions matter more than academic ones for compliance. Article 3(1) of the EU AI Act (Regulation (EU) 2024/1689) defines an AI system as a machine-based system designed to operate with varying levels of autonomy, that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers from the input it receives how to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments. The wording is aligned with the OECD definition revised in 2023. The key criterion is the capability to infer; Recital 12 and the Commission's 2025 guidelines on the definition exclude systems based solely on rules defined by natural persons to execute operations automatically. ISO/IEC 22989:2022 provides the corresponding vocabulary used by the ISO/IEC 42001 management-system standard.\n\nTwo recurring misconceptions: first, the distinction between narrow AI (competence on a defined task) and artificial general intelligence has no agreed test, and capability claims should be evaluated per task with benchmarks rather than by the label. Second, the so-called AI effect: once a technique works reliably (optical character recognition, spam filtering, route finding) people stop calling it AI, so the term tends to refer to whatever is currently new. For procurement and risk work the useful question is always what the system infers, from which data, and what decision it feeds.","da":"Begrebet blev skabt af John McCarthy i forslaget fra 1955 til Dartmouth Summer Research Project i 1956, som antog, at ethvert aspekt af læring eller intelligens i princippet kan beskrives så præcist, at en maskine kan simulere det. Alan Turing havde allerede i 1950 formuleret spørgsmålet operationelt med imitationsspillet. Fra starten delte feltet sig i to traditioner: symbolsk AI (logik, søgning, vidensrepræsentation, planlægning) og konnektionistisk AI baseret på netværk af simple enheder, der trænes på data. Symbolske metoder dominerede frem til 1980'erne med regelbaserede ekspertsystemer som højdepunkt; deres skrøbelighed og vedligeholdelsesomkostninger kombineret med overdrevne løfter bidrog til to perioder med faldende finansiering, de såkaldte AI-vintre (midten af 1970'erne og slutningen af 1980'erne).\n\nRussell og Norvig definerer feltet ud fra rationelle agenter: systemer, der opfatter et miljø og handler for at maksimere et forventet præstationsmål. Den ramme samler søgealgoritmer, constraint solvers, probabilistisk ræsonnement som bayesianske netværk, planlægning, robotteknologi og maskinlæring under ét. Siden 2010'erne er maskinlæring og især deep learning blevet den dominerende teknik, og derfor bruges AI og ML i daglig tale næsten som synonymer, og på det seneste også AI og store sprogmodeller. De er ikke det samme: en ruteplanlægger med A*-søgning eller en regelmotor til skatteberegning er AI i lærebogens forstand, men involverer ingen læring.\n\nI compliancearbejde vejer de juridiske definitioner tungere end de akademiske. Artikel 3, nr. 1, i AI-forordningen (forordning (EU) 2024/1689) definerer et AI-system som et maskinbaseret system, der er udformet til at fungere med varierende grader af autonomi, som kan udvise tilpasningsevne efter udrulning, og som til eksplicitte eller implicitte mål udleder af det input, det modtager, hvordan det skal generere output som forudsigelser, indhold, anbefalinger eller beslutninger, der kan påvirke fysiske eller virtuelle miljøer. Ordlyden er afstemt med OECD's definition, som blev revideret i 2023. Det afgørende kriterium er evnen til at udlede; betragtning 12 og Kommissionens retningslinjer fra 2025 om definitionen udelukker systemer, der alene bygger på regler fastsat af fysiske personer til automatisk udførelse af operationer. ISO/IEC 22989:2022 leverer det tilsvarende begrebsapparat, som ledelsessystemstandarden ISO/IEC 42001 bygger på.\n\nTo tilbagevendende misforståelser: For det første findes der ingen anerkendt test, der skiller snæver AI (kompetence inden for en afgrænset opgave) fra generel kunstig intelligens, så påstande om kunnen bør vurderes opgave for opgave med benchmarks frem for ud fra etiketten. For det andet den såkaldte AI-effekt: Når en teknik først virker pålideligt (tekstgenkendelse, spamfiltrering, rutefinding), holder man op med at kalde den AI, så ordet har tendens til at betegne det, der er nyt lige nu. I indkøb og risikovurdering er det nyttige spørgsmål altid, hvad systemet udleder, ud fra hvilke data, og hvilken beslutning det føder ind i."},"edges":[{"type":"contrasts-with","to":"ai/machine-learning","why":{"en":"AI is the whole goal of making computers act smart; machine learning is one way to get there, by learning from examples instead of written rules.","da":"AI er hele målet om at få computere til at handle klogt; maskinlæring er én vej dertil, hvor systemet lærer af eksempler i stedet for skrevne regler."},"confidence":"high","strength":"primary"},{"type":"contrasts-with","to":"ai/large-language-model","why":{"en":"Many people now say \"AI\" and mean a chat assistant, but a large language model is only one kind of AI among many.","da":"Mange siger i dag \"AI\" og mener en chatassistent, men en stor sprogmodel er kun én slags AI blandt mange."},"confidence":"high","strength":"normal"}],"depth":0,"sources":[{"title":"Russell & Norvig, Artificial Intelligence: A Modern Approach","url":"https://aima.cs.berkeley.edu/","tier":"textbook","publisher":"Pearson"},{"title":"ISO/IEC 22989:2022, Artificial intelligence concepts and terminology","url":"https://www.iso.org/standard/74296.html","tier":"standard","publisher":"ISO/IEC"},{"title":"Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 3(1) and Recital 12","url":"https://eur-lex.europa.eu/eli/reg/2024/1689/oj","tier":"standard","publisher":"European Union"},{"title":"NIST AI 100-1, Artificial Intelligence Risk Management Framework (AI RMF 1.0)","url":"https://doi.org/10.6028/NIST.AI.100-1","tier":"standard","publisher":"NIST"},{"title":"McCarthy, Minsky, Rochester & Shannon (1955), A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence","url":"http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html","tier":"reference"},{"title":"OECD (2023), Updates to the OECD's definition of an AI system explained","url":"https://oecd.ai/en/wonk/ai-system-definition-update","tier":"official-doc","publisher":"OECD"},{"title":"European Commission (2025), Guidelines on the definition of an artificial intelligence system","url":"https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-ai-system-definition-facilitate-first-ai-acts-rules-application","tier":"official-doc","publisher":"European Commission"}],"draft":true}