EVOPEDIA

← all systems · see it in The System

Artificial intelligence

Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making; it is also a field of research that develops and studies methods and software enabling machines to perceive their environment and act toward defined goals (src:wp:artificial-intelligence). In the field's agent frame, 'an agent is anything that perceives its environment through sensors and acts upon that environment through actuators' (src:wp:intelligent-agent). No definition of intelligence is agreed (C-1112).

Properties published scienceour proposal

Sourced fields, and our own readings.

born at1956 (Dartmouth workshop; proposal 31 August 1955; Logic Theorist presented there)
born precisionyear
boundaryInside: AI systems at work (machines that perceive through sensors, learn a model, search and reason, pursue an objective and act through actuators or outputs: deep-learning models, knowledge-based systems, reinforcement-learning agents, robots, automated vehicles, generative models) and the research field that builds, trains, benchmarks and studies them. Outside: the computers they run on (carrier), the people who label, use and fund them, the data on the Web, the electricity they draw, the brains they are compared with. The boundary depends on which definition of AI is used (C-1112).
nested levelRuns on computers (rung 12) as programs; built from algorithms, artificial neural networks and knowledge bases; the field sits inside science; deployed AI is embedded in human society and the economy. LLMs and LLM-based agents are kinds, walked on their own.
scaletechnological

Measures (22) published science

characteristic clock: training compute doubling time, deep-learning eraabout 6 (5.7 for regular-scale models, 2010-2022) months
training compute doubling time before 2010about 20 (21.3, 1952-2010) months
AlphaZero: random play to superhumanwithin 24 hours
major AI wintersabout 1974-1980 and 1987-2000 years
springs-to-winter lag (McDermott's warning to the next winter)about 5 years
the field's founding1956 (Dartmouth workshop; proposal dated 31 August 1955) year
definitions of intelligence collectedabout 70 definitions
ImageNet top-5 human error (annotator A1)5.1 %
DQN games above a human expert3 of 7 (and above all earlier approaches on 6 of 7) Atari 2600 games
AlphaGo vs Lee Sedol (March 2016)4 of 5 games won
visuomotor policy size92,000 parameters
wav2vec 2.0 word error rate with ten minutes of labels4.8 / 8.2 (clean / other), after 53k hours unlabelled pre-training % WER
CO2e of one NLP model development run with neural architecture search626,155 (vs 126,000 for an average car's lifetime incl. fuel) lbs CO2e
energy per chatbot query vs web searchabout 10 times
peak FLOPS vs DRAM vs interconnect bandwidth growth3.0x vs 1.6x vs 1.4x per 2 years (over 20 years)
speculative decoding speed-up2-3 times (T5-XXL, identical outputs)
Switch Transformer dropped (overflowed) tokenstypically < 1 % of tokens
GPipe bubble negligible whenM >= 4 x K (micro-batches vs partitions) ratio
benchmarks mapped3765 benchmarks across 947 tasks (computer vision and NLP)
model collapse: perplexity after training on generated data (5 epochs, no original data kept)losing performance from 20 to 28 perplexity points (base model 34 mean perplexity)
experts' 50% date for AI outperforming humans in all tasks (2016 survey)45 years years from the survey
US workforce with >=10% of tasks exposed to LLMsabout 80 (19% with >=50%)

Open contentions (8) open contention

Where sources or readings disagree. Both stay on the map.

Partners and parts published science

Built from / builds (11)

Sits inside / holds (3)

Trades with (valence) (26)

expressed by (14)

floor filled by (24)

represents (2)

Kinds (12)

Notes published science

Layers

The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.

percepts in, learned model, actions out (reader's view)

  1. L1: The environment the agent's sensors touch: light, sound, text, objects and surfaces, its own body, rewards from the environment
  2. L2: Sensors: cameras, microphones, lidar, radar, ultrasonic and tactile sensors, joint encoders; the reward channel; labelled data
  3. L3: Perception: deducing aspects of the world from sensor input (computer vision, speech recognition)
  4. L4: The learned model or knowledge base: weights of a neural network, or rules and facts
  5. L5: Inference: search, reasoning, planning
  6. L6: The objective: utility, reward or loss; goals held in an objective function, set by design
  7. L7: Action back on the world: the action expected to maximise the performance measure, taken through actuators or handed out as content and decisions

the boardroom (shared floors)

  1. L4: The learned model or knowledge: the weights of an artificial neural network, where meaning sits in the space of high-layer activations, or a knowledge base of rules and facts a program uses
  2. L5: Inference: search through states, reasoning, planning; step-by-step methods meet a combinatorial explosion; sparse models route each token to experts
  3. L6: The objective: an objective function, utility or reward that holds the goals, set by design together with the learning rule and architecture; the field's benchmarks steer which objectives are chosen
  4. L7: The action selected: the one expected to maximise the performance measure, given percepts, prior knowledge and available actions; booked in the field's mechanism words (C-432: nothing about understanding or meaning ruled; C-434: nothing about consciousness)

the field: benchmarks, results, recognition and funding (the research community)

  1. L1: Benchmarks: datasets and metrics
  2. L2: Benchmarks define tasks and steer progress
  3. L3: Results reported and recognised; paper acceptance
  4. L4: Funding and investment: springs and winters
  5. L5: The research agenda: the traditional goals (learning, reasoning, knowledge representation, planning, language, perception, robotics)

way in: Benchmark results (the field's channel)

  1. L1: Benchmarks: datasets and metrics that define tasks
  2. L2: Results reported on benchmarks; state-of-the-art (SOTA) curves
  3. L3: Recognition: models with new SOTA results receive widespread recognition
  4. L4: Scientific reward systems: paper acceptance tied to SOTA metrics
  5. L5: Enthusiasm, funding and jobs follow apparent breakthroughs, and dry up when progress stalls

way in: Compute and energy (electricity into accelerators)

  1. L1: Electricity: hardware, electricity and cloud compute paid for
  2. L2: Graphics processing units and other accelerators
  3. L3: Training compute (operations) spent on a model; memory bandwidth now the main bottleneck

way in: Hearing (microphones, speech audio)

  1. L1: Sound in air: speech and other audio
  2. L2: Microphone: the speech waveform
  3. L3: Latent speech representations learned from audio, masked and quantised

way in: Labels and human judgements (supervision)

  1. L1: People's answers: expert annotators and crowds labelling examples
  2. L2: Labelled training data: each example paired with its expected answer
  3. L3: The error between output and label, passed back along credit-assignment paths

way in: Own experience: games played against itself (inner channel)

  1. L1: Games the system played against itself, from random play
  2. L2: Self-play games recorded as experience for learning (tabula rasa reinforcement learning)
  3. L3: Learning by running experiments on itself

way in: Proprioception and touch (the machine's own body: joint encoders, vehicle state, tactile sensors)

  1. L1: The machine's own body state: joint angles, end-effector pose, the vehicle's own motion
  2. L2: Joint encoders; proprioceptive sensors for internal vehicle state; tactile sensors
  3. L3: Encoder readings fed with image features into the policy network

way in: Range sensing (lidar, radar, ultrasonic)

  1. L1: The environment's objects: dynamic and static, drivable areas, buildings, pedestrian crossings
  2. L2: Active sensors: lidar, radar, ultrasonic (they emit and read back)
  3. L3: Perception: raw sensor data turned into a description of the world (machine perception)

way in: Reward (how good the situation is)

  1. L1: The environment: what causes reward (a game won, a puzzle solved); the reward is part of the perception the environment generates
  2. L2: The reward signal: the simplest possible semantics
  3. L3: A value function estimating future rewards (Q-learning)

way in: Text (written language)

  1. L1: Written human language
  2. L3: Word embeddings: words as vectors encoding their meaning; transformers with attention

way in: Training data (the slow channel)

  1. L1: Huge datasets scraped from the Web
  2. L2: Curated datasets and benchmarks: millions of images in hundreds of categories
  3. L3: Learning from a stream of data: finding patterns without other guidance (unsupervised), or with labels

way in: Vision (cameras, images)

  1. L1: Light from the scene: colour and shape of objects, read passively
  2. L2: Camera: an image of raw pixels
  3. L3: Convolutional layers: a hierarchy of learned feature maps; intermediate layers track V4, the top layer IT in the primate ventral stream

way out: Classifications, rankings and decisions handed to people

  1. L1: Search results, assistants' answers and decisions people rely on; trust needs an explanation
  2. L7: The class predicted, the result ranked, the choice of maximum expected utility

way out: Generated content (text, images, audio, video)

  1. L1: Text, images, audio and video put into the world (and onto the Web)
  2. L7: A generator samples from the learned data distribution

way out: Heat and carbon (energy spent)

  1. L1: Carbon emitted to fuel the hardware; heat to the surroundings
  2. L4: Training a model consumes energy in proportion to its compute

way out: Motor commands (robots and vehicles)

  1. L1: Objects moved: screwing a cap onto a bottle; a vehicle driven
  2. L2: Torques at the joints; actuators
  3. L7: The policy picks the action

way out: Moves in a game or simulated world

  1. L1: Moves on the board or the game screen: a world-champion program defeated
  2. L7: A move chosen from the learned value of the positions

way out: The field's output: new benchmarks, results and predictions

  1. L1: Papers, benchmarks and confident predictions into the record
  2. L6: Novel benchmarks created to complement or replace saturated ones

Sources (48) published science

Artificial intelligence (Wikipedia, via the Wikimedia Enterprise snapshot of 2026-11-01)
Legg & Hutter, Universal Intelligence: A Definition of Machine Intelligence (Minds and Machines 2007; arXiv:0712.3329) (behind its links)
Silver et al., Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm (arXiv:1712.01815, 2017) (behind its links)
Mitchell et al., Orbital forcing of ice sheets during snowball Earth, Nature Communications 12, 4187 (2021), doi:10.1038/s41467-021-24439-4 (behind its links)
Ott et al., Mapping global dynamics of benchmark creation and saturation in artificial intelligence, Nature Communications 13:6793 (2022), doi:10.1038/s41467-022-34591-0 (CC BY; Europe PMC PMC9649641) (behind its links)
Sevilla et al., Compute Trends Across Three Eras of Machine Learning (arXiv:2202.05924, 2022) (behind its links)
Shumailov et al., AI models collapse when trained on recursively generated data, Nature 631:755-759 (2024), doi:10.1038/s41586-024-07566-y (CC BY; Europe PMC PMC11269175) (behind its links)
Goodfellow et al., Generative Adversarial Networks (arXiv:1406.2661, 2014) (behind its links)
Hassabis, Kumaran, Summerfield & Botvinick, Neuroscience-Inspired Artificial Intelligence, Neuron 95:245-258 (2017), doi:10.1016/j.neuron.2017.06.011 (behind its links)
Athalye, Carlini & Wagner, Obfuscated Gradients Give a False Sense of Security (ICML 2018; arXiv:1802.00420) (behind its links)
Eloundou et al., GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models (arXiv:2303.10130, 2023) (behind its links)
Varadi M, Bertoni D, Magana P, et al. (2024) AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences. Nucleic Acids Res 52(D1):D368-75. doi:10.1093/nar/gkad1011 (open access, PMC10767828) (behind its links)
McCarthy, Minsky, Rochester & Shannon, A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence (31 August 1955) (behind its links)
AI winter (Wikipedia, via the Wikimedia Enterprise snapshot of 2026-11-01) (behind its links)
Legg & Hutter, A Collection of Definitions of Intelligence (arXiv:0706.3639) (behind its links)
Russakovsky et al., ImageNet Large Scale Visual Recognition Challenge (IJCV 2015; arXiv:1409.0575) (behind its links)
Mnih et al., Playing Atari with Deep Reinforcement Learning (arXiv:1312.5602, 2013) (behind its links)
Levine, Finn, Darrell & Abbeel, End-to-End Training of Deep Visuomotor Policies (JMLR 2016; arXiv:1504.00702) (behind its links)
Baevski et al., wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations (NeurIPS 2020; arXiv:2006.11477) (behind its links)
Strubell, Ganesh & McCallum, Energy and Policy Considerations for Deep Learning in NLP (ACL 2019; arXiv:1906.02243) (behind its links)
Gholami et al., AI and Memory Wall (IEEE Micro 2024; arXiv:2403.14123) (behind its links)
Leviathan, Kalman & Matias, Fast Inference from Transformers via Speculative Decoding (ICML 2023; arXiv:2211.17192) (behind its links)
Fedus, Zoph & Shazeer, Switch Transformers (JMLR 2022; arXiv:2101.03961) (behind its links)
Huang et al., GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism (NeurIPS 2019; arXiv:1811.06965) (behind its links)
Grace et al., When Will AI Exceed Human Performance? Evidence from AI Experts (JAIR 2018; arXiv:1705.08807) (behind its links)
Yurtsever et al., A Survey of Autonomous Driving: Common Practices and Emerging Technologies (IEEE Access 2020; arXiv:1906.05113) (behind its links)
Intelligent agent (Wikipedia, via the Wikimedia Enterprise snapshot of 2026-11-01) (behind its links)
Lake, Ullman, Tenenbaum & Gershman, Building Machines That Learn and Think Like People (Behavioral and Brain Sciences 2017; arXiv:1604.00289) (behind its links)
Szegedy et al., Intriguing properties of neural networks (arXiv:1312.6199, 2013) (behind its links)
Schmidhuber, Deep Learning in Neural Networks: An Overview (Neural Networks 2015; arXiv:1404.7828) (behind its links)
Yamins et al., Performance-optimized hierarchical models predict neural responses in higher visual cortex, PNAS 111:8619-8624 (2014), doi:10.1073/pnas.1403112111 (behind its links)
Nightingale SJ, Farid H 2022. AI-synthesized faces are indistinguishable from real faces and more trustworthy. PNAS 119(8):e2120481119. doi:10.1073/pnas.2120481119 (behind its links)
Seff A, Ovadia Y, Zhou W, Adams RP (2020) SketchGraphs: A Large-Scale Dataset for Modeling Relational Geometry in Computer-Aided Design. arXiv:2007.08506 (behind its links)
Blanchard A, Taddeo M (2023) The Ethics of Artificial Intelligence for Intelligence Analysis: a Review of the Key Challenges with Recommendations. Digital Society, doi:10.1007/s44206-023-00036-4 (behind its links)
Dutilh Novaes, C. (2021, rev.) Argument and Argumentation. Stanford Encyclopedia of Philosophy (peer-reviewed reference work) (behind its links)
Wang Z, O'Boyle M 2018. Machine learning in compiler optimisation. Proceedings of the IEEE 106(11). arXiv:1805.03441 (behind its links)
Jobin A, Ienca M, Vayena E (2019). The global landscape of AI ethics guidelines. Nature Machine Intelligence 1:389-399. doi:10.1038/s42256-019-0088-2 (arXiv 1906.11668, preprint read) (behind its links)
Johnson CA, Willigan EM, Hwang EY, et al. (2026) Unlocking biodiversity data with robotic imaging and AI-driven transcription of natural history collections. Sci Adv. doi:10.1126/sciadv.aeb6975 (behind its links)
Computer science (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Lean (proof assistant) (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Epistemology (encyclopedia article, WME snapshot 2026-11-01) (behind its links)
Verschure PFMJ. The six capacities of responsible autonomy: naturalizing free will in biological and artificial systems. Phil Trans R Soc B 2026. doi:10.1098/rstb.2026.0001 (behind its links)
Yampolskiy RV. Human Indignity: From Legal AI Personhood to Selfish Memes. arXiv:1810.02724 (2018). (behind its links)
Tononi G., Koch C. (2015) Consciousness: here, there and everywhere? Phil Trans R Soc B 370:20140167. doi:10.1098/rstb.2014.0167 (behind its links)
Chollet F (2019) On the measure of intelligence. arXiv:1911.01547 (behind its links)
Legg S, Hutter M (2007) Universal intelligence: a definition of machine intelligence. Minds and Machines 17:391-444. arXiv:0712.3329 (behind its links)
Wang CX, Hilburn IA, Wu DA et al. (2019) Transduction of the Geomagnetic Field as Evidenced from alpha-Band Activity in the Human Brain. eNeuro 6(2). doi:10.1523/ENEURO.0483-18.2019 (PMC6494972) (behind its links)
Amodei et al., Concrete Problems in AI Safety (arXiv:1606.06565, 2016) (behind its links)