H
82 words and 231 other names, in italics, pointing to their entries.
- H+ → transhumanism
- H100
- H20 → H100
- H200 → H100
- H800 → H100
- Haar cascade → face detection
- Habsburg AI → model collapse
- HAI → Stanford HAI
- Hailuo AI → MiniMax (company)
- HAIP → G7 Hiroshima AI Process
- HAL → HAL 9000
- HAL 9000
- half → FP16
- half precision → FP16
- hallucinate → hallucination
- hallucinated dependency → slopsquatting
- hallucination
- halting problem → Turing machine
- Hamiltonian Monte Carlo → Markov chain Monte Carlo
- Hammersley–Clifford theorem → Markov random field
- hand off → handoff
- hand problem → AI hands
- handoff
- hand-off → handoff
- handwriting recognition → optical character recognition
- handwritten text recognition → optical character recognition
- Hangzhou DeepSeek Artificial Intelligence → DeepSeek
- Robin Hanson
- Harbor → Terminal-Bench
- hard ALife → artificial life
- hard margin → margin
- hard parameter sharing → multi-task learning
- hard problem → hard problem of consciousness
- hard problem of consciousness
- hard refusal → refusal
- hard takeoff → fast takeoff
- hard-exploration problem → sparse reward
- hardware FLOPs utilization → model FLOPs utilization
- hardware lottery
- hardware overhang → compute overhang
- hardware-enabled governance mechanism → on-chip governance mechanisms
- hardware-enabled mechanism → on-chip governance mechanisms
- harm classifier → safety classifier
- HarmBench
- harmful fine-tuning → fine-tuning attack
- harmless interpolation → benign overfitting
- harmlessness training → safety training
- harmonised standard → harmonised standards
- harmonised standards
- harmonium → restricted Boltzmann machine
- harmonized standards → harmonised standards
- harness
- harness engineering → harness
- Demis Hassabis
- hazard function → survival analysis
- HBM → high-bandwidth memory
- HBM bandwidth → memory bandwidth
- HBM3 → high-bandwidth memory
- HBM3E → high-bandwidth memory
- HBM4 → high-bandwidth memory
- HCAI → human-centered AI
- HCH
- HDBSCAN → DBSCAN
- HE → homomorphic encryption
- He initialization → weight initialization
- Kaiming He
- He Kaiming → Kaiming He
- head → attention head
- HealthBench
- HealthBench Consensus → HealthBench
- HealthBench Hard → HealthBench
- Heart on My Sleeve
- heavy-ball method → momentum
- Donald Hebb
- Hebb rule → Hebbian learning
- Hebbian learning
- Hebbian theory → Hebbian learning
- Hebb's postulate → Hebbian learning
- Hebb's rule → Hebbian learning
- held-out data → test set
- HellaSwag
- HELM
- HELM Capabilities → HELM
- HELM Lite → HELM
- HELM Safety → HELM
- helpful, honest and harmless
- helpful honest harmless → helpful, honest and harmless
- helpful-only → helpful-only model
- helpful-only model
- helpful-only variant → helpful-only model
- Helvetica scenario → mode collapse
- HEM → on-chip governance mechanisms
- hEN → harmonised standards
- Dan Hendrycks
- Hendrycks MATH → MATH
- Her (film)
- Herb Simon → Herbert A. Simon
- Herbert Alexander Simon → Herbert A. Simon
- Herbert Simon → Herbert A. Simon
- hero run → training run
- Hessian
- Hessian matrix → Hessian
- Hessian-free optimization → second-order optimization
- Hessian-vector product → Hessian
- hetero-associative memory → associative memory
- heterogeneous integration → advanced packaging
- heuristic
- Heuristic DENDRAL → DENDRAL
- heuristic evaluation function → evaluation function
- heuristic function → heuristic
- heuristic search → heuristic
- heuristics → heuristic
- hexapod → legged robot
- HF → Hugging Face
- HFU → model FLOPs utilization
- HG-DAgger → DAgger
- HHH → helpful, honest and harmless
- hidden activation → activation
- hidden feedback loop → feedback loop
- hidden layer
- hidden Markov model
- hidden scratchpad → scratchpad
- hidden size → hidden layer
- hidden state → hidden layer
- hidden technical debt → machine learning technical debt
- hidden unit → hidden layer
- hidden units → hidden layer
- hidden-state space → activation space
- hidden-state vector → activation vector
- hierarchical clustering
- hierarchical navigable small world → HNSW
- hierarchical navigable small world graph → HNSW
- hierarchical reasoning model
- hierarchical reinforcement learning
- hierarchical task network
- HiFi-GAN → vocoder
- high-bandwidth memory
- higher-order perception theory → higher-order theories of consciousness
- higher-order theories of consciousness
- higher-order theory → higher-order theories of consciousness
- higher-order thought theory → higher-order theories of consciousness
- High-level Advisory Body on Artificial Intelligence
- High-Level Expert Group on Artificial Intelligence → Ethics Guidelines for Trustworthy AI
- high-level machine intelligence
- highly reliable agent design → agent foundations
- High-NA EUV → extreme ultraviolet lithography
- high-quality data → data quality
- high-risk AI → high-risk AI system
- high-risk AI system
- high-risk system → high-risk AI system
- highway layer → highway network
- highway net → highway network
- highway network
- hill climbing
- hindsight experience replay → goal-conditioned reinforcement learning
- hinge loss
- Geoffrey Hinton
- HIP → ROCm
- HIPIFY → ROCm
- HiPPO → state space model
- Hiroshima AI Process → G7 Hiroshima AI Process
- Hiroshima Code of Conduct → G7 Hiroshima AI Process
- Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems → G7 Hiroshima AI Process
- HiSilicon → Huawei
- hit rate → recall
- hit rate@k → recall@k
- HITL → human-in-the-loop
- hitting a wall → scaling wall
- HLAB-AI → High-level Advisory Body on Artificial Intelligence
- HLAI → human-level AI
- HLE → Humanity's Last Exam
- HLE-Diamond → Humanity's Last Exam
- HLE-Rolling → Humanity's Last Exam
- HLMI → high-level machine intelligence
- HLO → XLA
- HMC → Markov chain Monte Carlo
- HMM → hidden Markov model
- HNSW
- Sepp Hochreiter
- Hoffmann scaling laws → Chinchilla scaling laws
- Douglas Hofstadter
- hold-out set → test set
- holdout set → validation set
- holdout test set → test set
- Holistic Evaluation of Language Models → HELM
- John Holland
- Holt's linear trend method → exponential smoothing
- Holt–Winters method → exponential smoothing
- homogenisation → homogenization
- homogenization
- homomorphic encryption
- honest AI → honesty
- honesty
- honey pot → honeypot
- honeypot
- honeytoken → honeypot
- H-only model → helpful-only model
- hook
- hooks → hook
- Hope (architecture) → nested learning
- John Hopfield
- Hopfield model → Hopfield network
- Hopfield net → Hopfield network
- Hopfield network
- Hopper → H100
- horizon effect → evaluation function
- horizontal scaling → scale-up and scale-out
- Horn clause
- Horn clauses → Horn clause
- Horn formula → Horn clause
- host processor → CPU
- HOT → higher-order theories of consciousness
- hotword → wake word
- how many r's in strawberry → strawberry problem
- Howard's algorithm → policy iteration
- HPE → pose estimation
- HPO → hyperparameter tuning
- HRAD → agent foundations
- HRI → human–robot interaction
- HRL → hierarchical reinforcement learning
- HRM → hierarchical reasoning model
- HSI → humanist superintelligence
- HSTU → generative recommendation
- HTN → hierarchical task network
- HTN planning → hierarchical task network
- HTR → optical character recognition
- HTTP 429 → rate limit
- Jensen Huang
- Huang's law
- Huawei
- Huawei Technologies → Huawei
- Huber function → Huber loss
- Huber loss
- HuBERT → wav2vec
- Hubert L. Dreyfus → Hubert Dreyfus
- Evan Hubinger
- Hugging Face
- Hugging Face attack → Hugging Face incident
- Hugging Face breach → Hugging Face incident
- Hugging Face Hub → Hugging Face
- Hugging Face incident
- Hugging Face Spaces → Hugging Face
- Hugging Face Transformers → Transformers
- Hughes phenomenon → curse of dimensionality
- human action recognition → action recognition
- human authorship requirement → AI authorship
- human control → human oversight
- human evaluation
- human judgment → human evaluation
- human language technology → natural language processing
- human oversight
- human pose estimation → pose estimation
- human preference data → preference data
- human rating → human evaluation
- human uplift study → uplift study
- human value alignment → value alignment
- human verification challenge → CAPTCHA
- human-centered AI
- human-centered artificial intelligence → human-centered AI
- human-centred AI → human-centered AI
- human-centric AI → human-centered AI
- HumanEval
- human-in-command → human oversight
- human-in-the-loop
- humanist superintelligence
- Humanity's Last Exam
- human-level AI
- human-level artificial intelligence → human-level AI
- human-level machine intelligence → human-level AI
- humanoid → humanoid robot
- humanoid robot
- human-out-of-the-loop → human-in-the-loop
- human–robot collaboration → collaborative robot
- human–robot interaction
- Humans Consulting HCH → HCH
- Marcus Hutter
- hybrid AI → neuro-symbolic AI
- hybrid architecture
- hybrid bonding → advanced packaging
- hybrid model → hybrid architecture
- hybrid parallelism → 3D parallelism
- hybrid reasoning → hybrid reasoning model
- hybrid reasoning model
- hybrid retrieval → hybrid search
- hybrid search
- hybrid thinking model → hybrid reasoning model
- hybrid transformer → hybrid architecture
- HyDE
- Hyena → state space model
- hype cycle
- hyperautomation → robotic process automation
- Hyperband
- hyperbolic tangent → tanh
- HyperNEAT → neuroevolution
- hypernet → hypernetwork
- hypernetwork
- HyperNetworks → hypernetwork
- hyperparameter
- hyper-parameter → hyperparameter
- hyperparameter optimization → hyperparameter tuning
- hyperparameter search → hyperparameter tuning
- hyperparameter sweep → hyperparameter tuning
- hyperparameter tuning
- hyperparameters → hyperparameter
- hyperscale cloud provider → hyperscaler
- hyperscaler
- hyperscalers → hyperscaler
- hypothesis class → hypothesis space
- hypothesis set → hypothesis space
- hypothesis space
- hypothesis testing
- hypothetical document embedding → HyDE
- Hypothetical Document Embeddings → HyDE