P
164 words and 526 other names, in italics, pointing to their entries.
- p50 latency → latency
- p99 latency → latency
- P(A|B) → conditional probability
- PAC → PAC learning
- PAC learnability → PAC learning
- PAC learning
- PAC model → PAC learning
- PAC-Bayes → PAC learning
- package hallucination → slopsquatting
- Pact for the Future → Global Digital Compact
- padding → stride
- padding mask → attention mask
- paged attention → PagedAttention
- paged KV cache → PagedAttention
- PagedAttention
- PAI → Partnership on AI
- PAIR
- PAIR attack → PAIR
- PAIR jailbreak → PAIR
- pairwise preferences → preference data
- pairwise ranking → learning to rank
- pairwise win rate → win rate
- PAL → program-aided language model
- PaliGemma → Gemma
- PaLM
- PaLM 2 → PaLM
- PaLM-E → PaLM
- pandas
- pandas DataFrame → pandas
- panoptic quality → panoptic segmentation
- panoptic segmentation
- panprotopsychism → panpsychism
- panpsychism
- panpsychist → panpsychism
- paperclip maximiser → paperclip maximizer
- paperclip maximizer
- paperclipper → paperclip maximizer
- Paperclips (game) → Universal Paperclips
- Papers with Code
- paperswithcode.com → Papers with Code
- Seymour Papert
- Parallel Distributed Processing
- parallel function calling → function calling
- parallel training → distributed training
- Parallel WaveNet → WaveNet
- parameter
- parameter count
- parameter decomposition
- parameter initialization → weight initialization
- parameter server
- parameter sharing → weight sharing
- parameter sweep → grid search
- parameter-efficient fine-tuning
- parameter-efficient transfer learning → parameter-efficient fine-tuning
- parameters → parameter
- parameter-server architecture → parameter server
- parametric → parametric model
- parametric knowledge
- parametric memory → parametric knowledge
- parametric method → parametric model
- parametric model
- parametric ReLU → leaky ReLU
- parametric statistics → parametric model
- params → parameter
- parasocial → parasocial relationship
- parasocial interaction → parasocial relationship
- parasocial relationship
- para-social relationship → parasocial relationship
- parity problem → XOR problem
- PARRY
- parser → syntactic parsing
- parsimony → Occam's razor
- parsing → syntactic parsing
- partial → partial derivative
- partial dependence → partial dependence plot
- partial dependence function → partial dependence plot
- partial dependence plot
- partial derivative
- partial differentiation → partial derivative
- partial mode collapse → mode collapse
- partially observable Markov decision process
- partially observable MDP → partially observable Markov decision process
- partial-order planning → automated planning
- particle filter
- particle swarm → particle swarm optimization
- particle swarm optimization
- Partnership on AI
- Partnership on Artificial Intelligence to Benefit People and Society → Partnership on AI
- part-of-speech tagging
- Parzen window → density estimation
- Parzen–Rosenblatt window method → density estimation
- Pascal's mugging
- pass → epoch
- pass at k → pass@k
- pass@1 → pass@k
- passage embedding → text embedding
- pass-hat-k → pass@k
- pass@k
- past key values → KV cache
- patch embedding
- patch tokens → patch embedding
- patchification → patch embedding
- patchify → patch embedding
- Patchscope → Patchscopes
- Patchscopes
- PATE
- path patching
- path planning
- pathfinding → path planning
- path-integrated gradients → integrated gradients
- Pathways Language Model → PaLM
- pathwise derivative estimator → reparameterization trick
- pathwise gradient → reparameterization trick
- patience → early stopping
- pattern classification → pattern recognition
- pattern recognition
- pattern–verbalizer pair → prompt template
- Paul F. Christiano → Paul Christiano
- Paul J. Werbos → Paul Werbos
- pause → AI pause
- Pause Giant AI Experiments
- Pause Giant AI Experiments: An Open Letter → Pause Giant AI Experiments
- pause letter → Pause Giant AI Experiments
- PauseAI
- PauseAI Global → PauseAI
- PauseAI US → PauseAI
- PBE → program synthesis
- PCA → principal component analysis
- P-consciousness → phenomenal consciousness
- pCTR → click-through rate prediction
- PD disaggregation → disaggregated serving
- PDDL → Planning Domain Definition Language
- PDDL+ → Planning Domain Definition Language
- PDDL2.1 → Planning Domain Definition Language
- PDDL3 → Planning Domain Definition Language
- PDF → probability density function
- p(doom)
- pdoom → p(doom)
- PDP Research Group → Parallel Distributed Processing
- PDP volumes → Parallel Distributed Processing
- peak data → data wall
- peak of inflated expectations → hype cycle
- peaking phenomenon → curse of dimensionality
- Judea Pearl
- Pearson correlation → correlation
- Pearson's r → correlation
- PEFT → parameter-efficient fine-tuning
- pelican benchmark → pelican on a bicycle
- pelican on a bicycle
- pelican riding a bicycle → pelican on a bicycle
- pelican test → pelican on a bicycle
- penalized sampling → repetition penalty
- penalty term → regularization
- Penn Treebank
- Penn Treebank tagset → Penn Treebank
- Roger Penrose
- Penrose–Lucas argument → Lucas–Penrose argument
- perceive → perception
- perception
- perception module → perception
- perception stack → perception
- perceptron
- perceptron learning rule → perceptron
- Perceptron Mark I → Mark I Perceptron
- Perceptrons: An Introduction to Computational Geometry → Perceptrons (book)
- Perceptrons (book)
- perceptual loss
- perceptual reality monitoring → higher-order theories of consciousness
- Performer → linear attention
- periodic averaging → local SGD
- per-layer transcoder → transcoder
- permission mode → permission system
- permission system
- permissions → permission system
- permutation feature importance → feature importance
- permutation importance → feature importance
- permutation invariance → invariance
- perplexity
- Perplexity AI
- perplexity filtering → quality filtering
- persistent memory → long-term memory
- persona pattern → role prompting
- persona prompting → role prompting
- persona training → character training
- persona vector
- persona vectors → persona vector
- personal data → personally identifiable information
- personal information → personally identifiable information
- personal superintelligence
- personality rights → right of publicity
- personality training → character training
- personally identifiable information
- person-based predictive policing → predictive policing
- personhood credential
- personhood credentials → personhood credential
- perverse instantiation
- petaflop-day → petaflop/s-day
- petaFLOPS → FLOPS
- petaflop/s-day
- petaflop/s-days → petaflop/s-day
- PETL → parameter-efficient fine-tuning
- PETs → privacy-preserving machine learning
- PF-day → petaflop/s-day
- PFN → tabular foundation model
- pfs-day → petaflop/s-day
- PGD adversarial training → adversarial training
- PGD attack → projected gradient descent attack
- PGM → probabilistic graphical model
- phase splitting → disaggregated serving
- phased rollout → staged deployment
- PHC → personhood credential
- phenomenal character → phenomenal consciousness
- phenomenal consciousness
- phenomenality → phenomenal consciousness
- phi → integrated information theory
- Φ → integrated information theory
- phi coefficient → Matthews correlation coefficient
- φ coefficient → Matthews correlation coefficient
- philosophical behaviorism → behaviorism
- philosophical zombie
- philosophy of AI → philosophy of artificial intelligence
- philosophy of artificial intelligence
- phone-use agent → GUI agent
- photogrammetry → 3D reconstruction
- photonic accelerator → photonic computing
- photonic computing
- phrase grounding → visual grounding
- pHRI → human–robot interaction
- physical adversarial example → adversarial patch
- physical AI
- physical artificial intelligence → physical AI
- physical human–robot interaction → human–robot interaction
- Physical Intelligence
- physical stance → intentional stance
- physical symbol system → physical symbol system hypothesis
- physical symbol system hypothesis
- physicalism → mind–body problem
- physics engine → physics simulator
- physics simulator
- physics-informed machine learning → physics-informed neural network
- physics-informed neural network
- pi 0 → π0
- π (company) → Physical Intelligence
- PI controller → PID controller
- π0
- π0.5 → π0
- π0-FAST → π0
- piano mover's problem → motion planning
- PID control → PID controller
- PID controller
- PID loop → PID controller
- PII → personally identifiable information
- Pile → The Pile
- -pilled
- PIM → analog in-memory computing
- pinball loss → quantile regression
- PINN → physics-informed neural network
- PINNs → physics-informed neural network
- pipeline bubble → pipeline parallelism
- pipeline jungle → machine learning technical debt
- pipeline model parallelism → pipeline parallelism
- pipeline orchestration → machine learning pipeline
- pipeline parallelism
- π_ref → reference model
- Walter Pitts
- pivotal act
- pivotal event → pivotal act
- pivotal outcome → pivotal act
- pivotal process → pivotal act
- pix2pix → image-to-image translation
- PixelCNN → autoregressive image generation
- PixelRNN → autoregressive image generation
- pixel-wise classification → semantic segmentation
- pi-zero → π0
- place-based predictive policing → predictive policing
- plan-and-execute
- plan-and-execute agent → plan-and-execute
- planner → automated planning
- planner–executor pattern → plan-and-execute
- planning as satisfiability → automated planning
- Planning Domain Definition Language
- planning with a learned model → model-based reinforcement learning
- plan-then-execute → plan-and-execute
- plasticity loss → loss of plasticity
- plate notation → probabilistic graphical model
- platonic representation hypothesis
- Platt scaling → calibration
- playing the training game → training gaming
- playout → rollout
- PLD → Product Liability Directive
- PLM → base model
- pLSA → topic model
- plug and play language model
- Pluribus → Libratus
- PM → reward model
- PMF → probability distribution
- PMI → mutual information
- PMLR → JMLR
- PMM → post-market monitoring
- point cloud
- point set → point cloud
- point tracking → object tracking
- PointNet → point cloud
- pointwise convolution → depthwise separable convolution
- pointwise mutual information → mutual information
- pointwise ranking → learning to rank
- poisoning attack → data poisoning
- Poisson regression → generalized linear model
- Polanyi paradox → Polanyi's paradox
- Polanyi's paradox
- polarity detection → sentiment analysis
- policy
- policy entropy collapse → entropy collapse
- policy gradient
- policy gradient method → policy gradient
- policy gradient theorem → policy gradient
- policy improvement → policy iteration
- policy iteration
- Polyak momentum → momentum
- Polyak–Ruppert averaging → exponential moving average of weights
- polynomial fit → polynomial regression
- polynomial kernel → kernel trick
- polynomial regression
- polysemantic → polysemanticity
- polysemantic neuron → polysemanticity
- polysemanticity
- POMDP → partially observable Markov decision process
- pool-based active learning → active learning
- pooling
- pooling layer → pooling
- Pope Francis puffer coat → Balenciaga Pope
- Pope in a puffer jacket → Balenciaga Pope
- Popper (ILP system) → inductive logic programming
- population drift → data drift
- population mean → expected value
- population stability index → data drift
- Porter stemmer → stemming
- POS tagger → part-of-speech tagging
- POS tagging → part-of-speech tagging
- PoS tagging → part-of-speech tagging
- pose estimation
- position embedding → positional encoding
- position encoding → positional encoding
- position evaluation → evaluation function
- position interpolation → context extension
- positional embedding → positional encoding
- positional encoding
- position-wise feed-forward network → feed-forward network
- positive feedback loop → feedback loop
- positive manifold → g factor
- positive predictive value → precision
- positive predictive value parity → predictive parity
- post hoc explainability → post hoc explanation
- post hoc explanation
- post hoc interpretability → post hoc explanation
- posterior
- posterior distribution → posterior
- posterior mode → maximum a posteriori estimation
- posterior predictive distribution → posterior
- posterior probability → posterior
- posterior sampling → Thompson sampling
- posterior sampling for reinforcement learning → Thompson sampling
- posting list → inverted index
- postings list → inverted index
- post-market monitoring
- post-market monitoring plan → post-market monitoring
- post-market monitoring system → post-market monitoring
- post-remote biometric identification → remote biometric identification
- post-train → post-training
- post-trained → post-training
- post-training
- posttraining → post-training
- post-training quantisation → post-training quantization
- post-training quantization
- PoT → program-aided language model
- potemkin → Potemkin understanding
- Potemkin understanding
- potemkins → Potemkin understanding
- potential outcome → counterfactual
- potential outcomes → causal inference
- potential-based reward shaping → reward shaping
- power concentration → concentration of power
- power law
- power usage effectiveness
- powered by AI → AI-powered
- powerful AI → country of geniuses in a datacenter
- power-law distribution → power law
- power-law relationship → power law
- power-law scaling → scaling laws
- power-seeking
- power-seeking AI → power-seeking
- power-seeking behavior → power-seeking
- PPLM → plug and play language model
- PPML → privacy-preserving machine learning
- PPO → proximal policy optimization
- PPO-Clip → proximal policy optimization
- PPO-Penalty → proximal policy optimization
- PPV → precision
- PPV parity → predictive parity
- PR curve → precision–recall curve
- precision
- precision matrix → precision
- precision@k → precision
- precision–recall curve
- pre-deployment evaluation → pre-deployment testing
- pre-deployment testing
- predeployment testing → pre-deployment testing
- predicate calculus → first-order logic
- predicate logic → first-order logic
- predict → prediction
- predicted click-through rate → click-through rate prediction
- predicted MOS → mean opinion score
- predicted value → prediction
- prediction
- prediction drift → model monitoring
- prediction error minimization → predictive processing
- prediction serving → model serving
- prediction set → conformal prediction
- prediction with expert advice → online learning
- predictions → prediction
- predictive AI
- predictive analytics → predictive AI
- predictive brain → predictive processing
- predictive coding → predictive processing
- predictive entropy → entropy
- predictive model → predictive AI
- predictive modeling → predictive AI
- predictive multiplicity → Rashomon effect
- predictive parity
- predictive policing
- predictive processing
- predictive uncertainty → uncertainty quantification
- predictor → feature
- preference data
- preference dataset → preference data
- preference model → reward model
- prefill
- prefill attack
- prefill chunking → chunked prefill
- prefill jailbreak → prefill attack
- prefill phase → prefill
- prefill stage → prefill
- prefill–decode disaggregation → disaggregated serving
- prefilling → prefill
- prefilling attack → prefill attack
- prefix caching → prompt caching
- prefix LM → decoder-only transformer
- prefix tuning
- PReLU → leaky ReLU
- Preparedness Framework
- prequential code length → minimum description length
- pre-release testing → pre-deployment testing
- presence penalty → repetition penalty
- presumption of conformity → harmonised standards
- PreToolUse hook → hook
- pre-train → pretraining
- pretrain → pretraining
- pretrained language model → base model
- pretrained-only model → base model
- pretraining
- pre-training → pretraining
- pretraining compute → training compute
- pretraining corpus → pretraining data
- pretraining data
- pre-training data → pretraining data
- pretraining dataset → pretraining data
- pretraining mix → data mixture
- pretrain-then-fine-tune → transfer learning
- Preventing Woke AI in the Federal Government → woke AI
- previous-token head → induction head
- PRH → platonic representation hypothesis
- price per million tokens → inference cost
- primitive acts → conceptual dependency
- Princeton WordNet → WordNet
- principal component analysis
- principal components → principal component analysis
- principle of parsimony → Occam's razor
- prior
- prior box → anchor box
- prior distribution → prior
- prior probability → prior
- prior probability shift → label shift
- prior shift → label shift
- prior-data fitted network → tabular foundation model
- prioritized experience replay → experience replay
- prior-preservation loss → DreamBooth
- priors → prior
- privacy
- privacy accounting → privacy budget
- privacy as contextual integrity → contextual integrity
- privacy attack → privacy
- privacy budget
- privacy loss budget → privacy budget
- privacy parameter → privacy budget
- privacy-enhancing technologies → privacy-preserving machine learning
- privacy-preserving AI → privacy-preserving machine learning
- privacy-preserving machine learning
- Private Aggregation of Teacher Ensembles → PATE
- privilege levels → instruction hierarchy
- privileged basis
- probabilistic → probability
- probabilistic forecasting → time series forecasting
- probabilistic graphical model
- probabilistic latent semantic analysis → topic model
- probabilistic programming
- probabilistic programming language → probabilistic programming
- probabilistic roadmap
- probabilistic roadmap method → probabilistic roadmap
- probabilistic topic model → topic model
- probabilities → probability
- probability
- probability calibration → calibration
- probability density → probability density function
- probability density estimation → density estimation
- probability density function
- probability distribution
- probability flow ODE → score-based generative model
- probability mass function → probability distribution
- probability of doom → p(doom)
- probability score → Brier score
- probability theory → probability
- probability value → p-value
- probably approximately correct learning → PAC learning
- probe → probing
- probing
- probing classifier → probing
- probing task → probing
- problem decomposition → task decomposition
- problem of control → control problem
- problem of induction → inductive reasoning
- problem of other minds
- problem space → state space search
- procedural art → generative art
- Procedural Reasoning System → belief–desire–intention model
- Proceedings of Machine Learning Research → JMLR
- process node
- process reward model
- process supervision
- process verifier → process reward model
- process-based feedback → process supervision
- process-based supervision → process supervision
- processing-in-memory → analog in-memory computing
- processor → CPU
- processor–memory gap → memory wall
- process-supervised reward model → process reward model
- Product Liability Directive
- product quantization
- production monitoring → model monitoring
- production rule → production system
- production system
- productionization → model deployment
- production-rule system → production system
- productivity J-curve → productivity paradox
- productivity paradox
- Progol → inductive logic programming
- program of thoughts → program-aided language model
- program synthesis
- program-aided language model
- program-aided reasoning → program-aided language model
- programmatic labeling → weak supervision
- programmation en logique → Prolog
- programming by demonstration → imitation learning
- programming by example → program synthesis
- program-size complexity → Kolmogorov complexity
- progressive delivery → canary deployment
- progressive disclosure
- progressive distillation → diffusion distillation
- progressive sharpening → edge of stability
- prohibited AI practices
- Project Glasswing
- project instructions → agent instruction file
- Project Jupyter → Jupyter notebook
- Project MAC → MIT AI Lab
- Project RAND → RAND Corporation
- Project Vend → Vending-Bench
- projected gradient descent attack
- projection ablation → directional ablation
- Prolog
- prompt
- Prompt Automatic Iterative Refinement → PAIR
- prompt brittleness → prompt sensitivity
- prompt caching
- prompt chain → prompt chaining
- prompt chaining
- prompt crafting → prompt engineering
- prompt design → prompt engineering
- prompt designer → prompt engineer
- prompt distillation → context distillation
- prompt engineer
- prompt engineering
- prompt extraction → prompt leaking
- prompt format → chat template
- prompt fragility → prompt sensitivity
- prompt injection
- prompt injection attack → prompt injection
- prompt leakage → prompt leaking
- prompt leaking
- prompt management
- prompt optimization
- prompt pattern → prompt template
- prompt phase → prefill
- prompt processing → prefill
- prompt programming → prompt engineering
- prompt registry → prompt management
- prompt sensitivity
- prompt template
- prompt tuning
- prompt versioning → prompt management
- prompt worm → AI worm
- promptable concept segmentation → Segment Anything Model
- promptable segmentation → image segmentation
- prompting → prompt
- PromptOps → prompt management
- prompts → prompt
- proof assistant → interactive theorem proving
- proof of personhood → personhood credential
- proper orthogonal decomposition → principal component analysis
- proper scoring rule
- proportional–integral–derivative controller → PID controller
- propositional calculus → propositional logic
- propositional logic
- propositional satisfiability → Boolean satisfiability problem
- proprietary model
- proprioception
- proprioceptive → proprioception
- proprioceptive sensing → proprioception
- proprioceptive state → proprioception
- prosaic AGI → prosaic alignment
- prosaic AI alignment → prosaic alignment
- prosaic alignment
- protected attribute
- protected characteristic → protected attribute
- protected class → protected attribute
- protected ground → protected attribute
- ProtoPNet → prototypical part network
- prototype network → prototypical part network
- prototype-based model → prototypical part network
- prototypical part network
- provable robustness → certified robustness
- provable safety → guaranteed safe AI
- provably safe AI → guaranteed safe AI
- provenance → data provenance
- prover–verifier game
- prover–verifier games → prover–verifier game
- provider → provider and deployer
- provider and deployer
- proximal policy optimization
- proxy discrimination
- proxy gaming → specification gaming
- proxy variable → proxy discrimination
- prune → pruning
- pruning
- PS architecture → parameter server
- pseudo-Huber loss → Huber loss
- pseudo-labeling → semi-supervised learning
- PSI → data drift
- PSM → fill-in-the-middle
- PSO → particle swarm optimization
- PSSH → physical symbol system hypothesis
- psychologism → Blockhead
- psychology of AI → machine psychology
- psychometric g → g factor
- PTB → Penn Treebank
- PTD-P → 3D parallelism
- .pte → ExecuTorch
- PTQ → post-training quantization
- publicity rights → right of publicity
- PUCT → Monte Carlo tree search
- PUE → power usage effectiveness
- puffer Pope → Balenciaga Pope
- pure-play foundry → semiconductor foundry
- Putnam benchmark → PutnamBench
- PutnamBench
- p-value
- PwC → Papers with Code
- PWN → WordNet
- pyannote → speaker diarization
- PyMC → probabilistic programming
- Pyro → probabilistic programming
- PySR → symbolic regression
- PyTensor → Theano
- PyTorch
- PyTorch 2 → PyTorch
- PyTorch Edge → ExecuTorch
- PyTorch Foundation
- PyTorch Hub → model hub
- PyTorch Mobile → ExecuTorch
- pytorch-pretrained-bert → Transformers
- pytorch-transformers → Transformers
- p-zombie → philosophical zombie