S
257 words and 825 other names, in italics, pointing to their entries.
- S2S → speech-to-speech model
- S2ST → speech-to-speech model
- S4 → state space model
- S6 → Mamba
- SA → simulated annealing
- SA-1B → Segment Anything Model
- sabotage capability evaluation → sabotage evaluation
- sabotage eval → sabotage evaluation
- sabotage evaluation
- SAC → soft actor–critic
- SacreBLEU → BLEU
- SAD → voice activity detection
- saddle → saddle point
- saddle point
- saddle-point problem → saddle point
- SAE → sparse autoencoder
- SAE feature → feature (interpretability)
- SAE feature steering → feature steering
- SAE J3016 → levels of driving automation
- SAE latent → feature (interpretability)
- SAE levels → levels of driving automation
- SAE steering → feature steering
- SAEs → sparse autoencoder
- safe AI → AI safety
- Safe and Secure Innovation for Frontier Artificial Intelligence Models Act → SB 1047
- safe completion
- safe exploration → safe reinforcement learning
- safe interruptibility
- safe reinforcement learning
- safe RL → safe reinforcement learning
- Safe Superintelligence → Safe Superintelligence Inc.
- Safe Superintelligence Inc.
- safe-completion training → safe completion
- safely interruptible agent → safe interruptibility
- safetensors
- safetensors format → safetensors
- safety argument → safety case
- safety case
- safety classifier
- safety fine-tuning → safety training
- safety guardrails → guardrails
- safety post-training → safety training
- safety tax → alignment tax
- safety training
- safety tuning → safety training
- safety washing → safetywashing
- safetywashing
- safety-washing → safetywashing
- SAH → superficial alignment hypothesis
- saliency → saliency map
- saliency map
- Sally–Anne test → theory of mind
- SAM 2 → Segment Anything Model
- SAM 3 → Segment Anything Model
- Sam Altman's firing → OpenAI leadership crisis (2023)
- Samantha → Her (film)
- SAMME → AdaBoost
- sample bias → sampling bias
- sample efficiency
- sample rate → sampling
- sample selection bias → sampling bias
- sample size → sampling
- sampled → sampling
- sample-efficient → sample efficiency
- sampler → diffusion sampler
- samples → sampling
- sample-wise double descent → double descent
- sampling
- sampling bias
- sampling method → diffusion sampler
- sampling steps → diffusion sampler
- sampling strategy → decoding
- sampling temperature → temperature
- sampling-based planning → motion planning
- Samuel Altman → Sam Altman
- Arthur Samuel
- Samuel Checkers-playing Program → Samuel's checkers program
- Samuel Harris Altman → Sam Altman
- Samuel's checker player → Samuel's checkers program
- Samuel's checkers program
- sandbag → sandbagging
- sandbagging
- sandbox
- sandboxed → sandbox
- sandboxing → sandbox
- sandwich experiment → sandwiching
- sandwiching
- sandwiching experiment → sandwiching
- SARIMA → ARIMA
- SARSA
- SARSA(λ) → SARSA
- SAT → Boolean satisfiability problem
- SAT solver → Boolean satisfiability problem
- SAT solving → Boolean satisfiability problem
- satisfiability modulo theories
- satisfice → satisficing
- satisficer → satisficing
- satisficing
- satisficing agent → satisficing
- saturate → benchmark saturation
- saturated benchmark → benchmark saturation
- saturation → benchmark saturation
- SB 1047
- SB 24-205 → Colorado AI Act
- SB 243
- SB 26-189 → Colorado AI Act
- SB 53
- SBC → Bayesian information criterion
- SBERT → Sentence-BERT
- SC 42 → ISO/IEC JTC 1/SC 42
- scaffold → scaffolding
- scaffolded model → scaffolding
- scaffolding
- SCAI → seemingly conscious AI
- scalable oversight
- scalar
- scalar product → dot product
- scalar quantity → scalar
- scalar reward → reward
- scalar value → scalar
- Scale → Scale AI
- Scale AI
- scale is all you need → scaling hypothesis
- scale out → scale-up and scale-out
- scale up → scale-up and scale-out
- scale-across → scale-up and scale-out
- scaled dot-product attention
- scale-free distribution → power law
- scale-invariant feature transform
- scale-out → scale-up and scale-out
- scale-pilled → -pilled
- scale-up → scale-up and scale-out
- scale-up and scale-out
- scale-up domain → scale-up and scale-out
- scaling hypothesis
- scaling law → scaling laws
- scaling laws
- scaling maximalism → scaling hypothesis
- scaling RL → RL scaling
- scaling wall
- scaling with repeated data → data-constrained scaling
- scaling-pilled → -pilled
- SCARA robot → industrial robot
- scene parsing → semantic segmentation
- scene reconstruction → 3D reconstruction
- scene text recognition → optical character recognition
- Schedule-Free AdamW → schedule-free optimization
- schedule-free learning → schedule-free optimization
- schedule-free optimization
- Schedule-Free SGD → schedule-free optimization
- schema theorem → genetic algorithm
- schema-constrained output → structured outputs
- Scheme → Lisp
- schemer → scheming
- scheming
- Jürgen Schmidhuber
- John Schulman
- Schwarz criterion → Bayesian information criterion
- Schwarz information criterion → Bayesian information criterion
- SCI → Strategic Computing Initiative
- scikit-learn
- scikits.learn → scikit-learn
- SCM → structural causal model
- score distillation sampling
- score matching → score-based generative model
- score-based diffusion → score-based generative model
- score-based generative model
- score-based model → score-based generative model
- score-function estimator → policy gradient
- scoring rule → proper scoring rule
- scrape → web scraping
- scraper → web scraping
- scraping → web scraping
- scratch pad → scratchpad
- scratchpad
- screen scraping → web scraping
- script
- script theory → script
- scripts → script
- scrubbing → causal scrubbing
- scruffies → neats and scruffies
- scruffy AI → neats and scruffies
- SD → Stable Diffusion
- SD 1.5 → Stable Diffusion
- SDAE → denoising autoencoder
- SDC → silent data corruption
- SDD → spec-driven development
- SDEdit → image-to-image translation
- SDPA → scaled dot-product attention
- SDS → score distillation sampling
- SDXL → Stable Diffusion
- SE block → squeeze-and-excitation network
- seahorse emoji
- seahorse emoji meltdown → seahorse emoji
- search problem → state space search
- search space → state space search
- search tree → game tree
- John Searle
- SecAgg → secure aggregation
- second central moment → variance
- second incompleteness theorem → Gödel's incompleteness theorems
- second-order approximation → Taylor series
- second-order method → second-order optimization
- second-order optimization
- second-stage ranking → reranking
- Section 230
- Section 230 of the Communications Decency Act → Section 230
- secure aggregation
- secure aggregation protocol → secure aggregation
- secure computation → secure multi-party computation
- secure enclave → trusted execution environment
- secure multi-party computation
- secure multiparty computation → secure multi-party computation
- securing model weights → model weight security
- security level (SL1–SL5) → model weight security
- security of AI → AI security
- SED → sound event detection
- See5 → C4.5
- seed → random seed
- seed AI
- seed artificial intelligence → seed AI
- seed parameter → nondeterminism
- seed value → random seed
- seemingly conscious AI
- Segment Anything → Segment Anything Model
- Segment Anything Model
- segmentation → image segmentation
- segmentation mask → image segmentation
- segment-level recurrence → Transformer-XL
- Terry Sejnowski
- selective answering → abstention
- selective prediction → abstention
- selective SSM → Mamba
- selective state space model → Mamba
- selectivity → specificity
- self-attention
- self-attention layer → self-attention
- self-aware → self-awareness
- self-awareness
- self-consciousness → self-awareness
- self-consistency
- self-consistency decoding → self-consistency
- self-correct → self-correction
- self-correction
- self-debugging → self-correction
- self-distillation → knowledge distillation
- self-distillation with no labels → DINO
- self-driving car → autonomous vehicle
- self-driving car dilemma → trolley problem
- self-driving taxi → robotaxi
- self-exfiltrate → self-exfiltration
- self-exfiltration
- self-generated instructions → Self-Instruct
- self-improve → self-improvement
- self-improvement
- self-improving AI → recursive self-improvement
- self-improving model → self-improvement
- Self-Instruct
- Self-Instruct method → Self-Instruct
- self-knowledge → self-awareness
- self-model → self-awareness
- self-normalized importance sampling → importance sampling
- self-organising map → self-organizing map
- self-organizing feature map → self-organizing map
- self-organizing map
- self-paced learning → curriculum learning
- self-play
- self-proliferation → autonomous replication and adaptation
- self-refine → self-correction
- self-refinement → self-correction
- self-repair → self-correction
- self-replicating prompt → AI worm
- self-report → introspection
- Oliver Selfridge
- self-stabilization → edge of stability
- self-supervised learning
- self-supervised pretraining → self-supervised learning
- self-supervised speech representation → wav2vec
- self-supervision → self-supervised learning
- self-taught reasoner
- semantic cache → semantic caching
- semantic caching
- semantic chunking → chunking
- semantic entropy
- Semantic ID → generative recommendation
- Semantic Kernel → AutoGen
- semantic net → semantic network
- semantic network
- semantic parsing
- semantic search
- semantic segmentation
- semantic textual similarity
- semantic uncertainty → semantic entropy
- semantic web
- semiconductor export controls → export controls on AI chips
- semiconductor foundry
- semi-structured sparsity → N:M sparsity
- semi-supervised learning
- semisupervised learning → semi-supervised learning
- SENet → squeeze-and-excitation network
- sense disambiguation → word sense disambiguation
- sensitive attribute → protected attribute
- sensitivity → recall
- sensitivity map → saliency map
- sensor fusion
- sentence embedding → text embedding
- sentence encoder → embedding model
- sentence similarity → semantic textual similarity
- sentence transformer → embedding model
- Sentence Transformers → Sentence-BERT
- Sentence-BERT
- SentencePiece
- sentence-transformers → Sentence-BERT
- sentential logic → propositional logic
- sentience
- sentient → sentience
- sentiment analysis
- sentiment classification → sentiment analysis
- sentinel token → special token
- Seoul commitments → Frontier AI Safety Commitments
- Seoul Declaration → AI Safety Summit
- Seoul Frontier AI Safety Commitments → Frontier AI Safety Commitments
- separable convolution → depthwise separable convolution
- separating hyperplane → decision boundary
- separation (fairness criterion) → equalized odds
- seq2seq → sequence-to-sequence model
- sequence parallelism
- sequence to sequence → sequence-to-sequence model
- sequence transduction model → sequence-to-sequence model
- sequence-parallel training → sequence parallelism
- sequence-to-sequence model
- sequential learning → continual learning
- sequential model-based optimization → Bayesian optimization
- sequential Monte Carlo → particle filter
- series elastic actuator → actuator
- serious incident
- serious incident reporting → serious incident
- server-sent events → streaming
- service-level agreement → service-level objective
- service-level indicator → service-level objective
- service-level objective
- serving → model serving
- serving cost → inference cost
- servo → actuator
- servomotor → actuator
- sexual deepfake → non-consensual intimate imagery
- SfM → structure from motion
- SFT → supervised fine-tuning
- SFT data → instruction dataset
- SFT stage → supervised fine-tuning
- SGD → stochastic gradient descent
- SGD with momentum → momentum
- SGDR → cosine annealing
- SGLang
- SGNS → word2vec
- shadow deployment
- shadow mode → shadow deployment
- shadow testing → shadow deployment
- Shakey → Shakey the robot
- Shakey the robot
- Shampoo
- Claude Shannon
- Shannon entropy → entropy
- Shannon theory → information theory
- SHAP
- SHAP values → SHAP
- shaped reward → reward shaping
- shaping → reward shaping
- SHapley Additive exPlanations → SHAP
- Shapley value
- Shapley values → Shapley value
- shard → shard theory
- shard theory
- shard theory of human values → shard theory
- sharded data parallelism → fully sharded data parallel
- shared expert → mixture of experts
- shared task → task (machine learning)
- shared weights → weight sharing
- sharp and flat minima
- sharp left turn
- sharp minima → sharp and flat minima
- sharp minimum → sharp and flat minima
- sharp takeoff → fast takeoff
- sharpness → sharp and flat minima
- sharpness-aware minimization
- shattered gradients → gradient masking
- shattering → VC dimension
- Noam Shazeer
- ShieldGemma → Gemma
- SHIFT → sparse feature circuit
- Buck Shlegeris
- shoggoth
- shoggoth meme → shoggoth
- shoggoth with a smiley face → shoggoth
- shooting method → trajectory optimization
- SHOP2 → hierarchical task network
- short timelines → AI timelines
- short-circuiting → circuit breakers
- shortcut connection → residual connection
- shortcut learning
- short-term memory → working memory
- short-time Fourier transform → spectrogram
- SHRDLU
- shrinkage → regularization
- Shunichi Amari → Shun-ichi Amari
- shutdown mechanism → kill switch
- shutdown problem
- shutdown resistance → shutdown problem
- SI → swarm intelligence
- SIAI → Machine Intelligence Research Institute
- Siamese network
- Siamese neural network → Siamese network
- side effect → negative side effects
- side effects → negative side effects
- side-effect penalty → impact measure
- SIFT → scale-invariant feature transform
- SIFT descriptor → scale-invariant feature transform
- SIFT keypoint → scale-invariant feature transform
- SIGKDD → KDD
- SigLIP → CLIP
- σ → variance
- σ² → variance
- sigmoid → sigmoid function
- sigmoid function
- sigmoid linear unit → SiLU
- sigmoid-weighted linear unit → SiLU
- signal temporal logic → temporal logic
- significance test → hypothesis testing
- silent data corruption
- silent data error → silent data corruption
- silent deployment → shadow deployment
- silent error → silent data corruption
- silhouette → silhouette score
- silhouette coefficient → silhouette score
- silhouette plot → silhouette score
- silhouette score
- silhouette width → silhouette score
- silicon photonics → photonic computing
- SiLU
- David Silver
- sim2real → sim-to-real transfer
- similarity learning → metric learning
- similarity search → vector search
- SimMIM → masked autoencoder
- Herbert A. Simon
- simple Bayes → naive Bayes
- simple exponential smoothing → exponential smoothing
- simple linear regression → linear regression
- simple recurrent network → recurrent neural network
- simple regret → regret
- SimpleQA
- SimpleQA Verified → SimpleQA
- simplex method → linear programming
- simplicity bias → Occam's razor
- Simpson paradox → Simpson's paradox
- Simpson's paradox
- sim-to-real → sim-to-real transfer
- sim-to-real gap → sim-to-real transfer
- sim-to-real transfer
- simulacra → simulator theory
- simulacrum → simulator theory
- simulated annealing
- simulated quantization → quantization-aware training
- simulation environment → physics simulator
- simulation-to-reality transfer → sim-to-real transfer
- simulator framing → simulator theory
- simulator theory
- simulators → simulator theory
- simultaneous localization and mapping
- SINDy → symbolic regression
- single precision → FP32
- single-image super-resolution → super-resolution
- single-layer perceptron → perceptron
- single-object tracking → object tracking
- single-source shortest path → Dijkstra's algorithm
- single-stage detector → object detection
- singleton
- singular learning theory
- singular value → singular value decomposition
- singular value decomposition
- singular values → singular value decomposition
- singular vectors → singular value decomposition
- singularitarian → technological singularity
- singularitarianism → technological singularity
- singularity → technological singularity
- Singularity Institute → Machine Intelligence Research Institute
- Singularity Institute for Artificial Intelligence → Machine Intelligence Research Institute
- sink token → attention sink
- Sinkhorn distance → Wasserstein distance
- sinusoidal positional encoding → positional encoding
- SIREN → implicit neural representation
- SISA training → machine unlearning
- situated cognition → embodied cognition
- situation calculus
- situational awareness
- Situational Awareness (essay)
- Situational Awareness: The Decade Ahead → Situational Awareness (essay)
- situationally aware → situational awareness
- situationally aware model → situational awareness
- six-fingered hands → AI hands
- six-month pause letter → Pause Giant AI Experiments
- skeleton tracking → pose estimation
- skewed classes → class imbalance
- skill → agent skill
- skill atrophy → deskilling
- skill erosion → deskilling
- SKILL.md → agent skill
- skip connection → residual connection
- skip transcoder → transcoder
- skip-gram → word2vec
- skip-gram with negative sampling → word2vec
- sklearn → scikit-learn
- Skynet
- SLA → service-level objective
- SLAM → simultaneous localization and mapping
- SLD resolution → resolution
- Sledgehammer → Isabelle
- sleeper agent
- sleeper agents → sleeper agent
- sleep-time compute
- SLI → service-level objective
- sliding window attention
- SLM → small language model
- SLO → service-level objective
- slop → AI slop
- slop words → LLM-isms
- slopification → AI slop
- slopsquatting
- slot filling → natural language understanding
- slot-filler representation → frame
- slow node → straggler
- slow takeoff
- SLT → singular learning theory
- small AI
- small language model
- small model → small language model
- Smallville → generative agent
- smart assistant → virtual assistant
- smart surveillance → AI surveillance
- SMBO → Bayesian optimization
- SMC → particle filter
- smooth grad → SmoothGrad
- smooth L1 loss → Huber loss
- smoothed classifier → randomized smoothing
- smoothed gradients → SmoothGrad
- SmoothGrad
- SmoothQuant
- SMOTE
- SMOTE-NC → SMOTE
- SMPC → secure multi-party computation
- SMT solver → satisfiability modulo theories
- SMT solving → satisfiability modulo theories
- SMT-LIB → satisfiability modulo theories
- SNE → t-SNE
- SNN → spiking neural network
- so over, so back → it's over, we're so back
- Soar → cognitive architecture
- social chatbot → AI companion
- social credit scoring → social scoring
- social robot → human–robot interaction
- social robotics → human–robot interaction
- social score → social scoring
- social scoring
- Society of Mind → The Society of Mind
- society of mind theory → The Society of Mind
- soft actor–critic
- soft ALife → artificial life
- soft attention → attention
- soft computing
- soft labels → label smoothing
- soft margin → support vector machine
- soft optimization → mild optimization
- soft parameter sharing → multi-task learning
- soft prompt → prompt tuning
- soft prompt tuning → prompt tuning
- soft prompting → prompt tuning
- soft takeoff → slow takeoff
- soft target update → target network
- softmax
- softmax cross-entropy → cross-entropy
- softmax function → softmax
- softmax normalizer loss → z-loss
- softmax regression → logistic regression
- softmax temperature → temperature
- soft-NMS → non-maximum suppression
- Software 2.0
- Software 3.0
- software bot → robotic process automation
- software engineering agent → coding agent
- software factory → dark factory
- software intelligence explosion → intelligence explosion
- software progress → algorithmic progress
- software robot → robotic process automation
- Jascha Sohl-Dickstein
- solely automated decision → automated decision-making
- SolidGoldMagikarp → glitch token
- Solomonoff induction
- Solomonoff prior → Solomonoff induction
- Ray Solomonoff
- Solomonoff's theory of inductive inference → Solomonoff induction
- Solow paradox → productivity paradox
- Solow's paradox → productivity paradox
- solutionism → techno-solutionism
- SOM → self-organizing map
- sonogram → spectrogram
- Sora
- Sora 2 → Sora
- Sora app → Sora
- Sora Turbo → Sora
- SORT → object tracking
- SOTA → state of the art
- SoTA → state of the art
- SOTA-chasing → state of the art
- sound event detection
- SoundStream → neural audio codec
- source attribution → citation
- source domain → domain adaptation
- source separation
- sovereign AI
- SP → sequence parallelism
- space data center → orbital data center
- spacetime patches → diffusion transformer
- SpaceXAI → xAI
- spaghetti test → Will Smith eating spaghetti
- span corruption → T5
- Sparks of AGI
- Sparks of Artificial General Intelligence → Sparks of AGI
- Sparks paper → Sparks of AGI
- SPARQL → Resource Description Framework
- sparse → sparsity
- sparse attention
- sparse autoencoder
- sparse coding → dictionary learning
- sparse crosscoder → crosscoder
- sparse dictionary learning → dictionary learning
- sparse feature circuit
- sparse feature circuits → sparse feature circuit
- sparse matrix → sparsity
- sparse mixture of experts → mixture of experts
- sparse model → sparsity
- sparse retrieval
- sparse reward
- sparse rewards → sparse reward
- Sparse Transformer → sparse attention
- sparse upcycling
- sparsely gated mixture of experts → mixture of experts
- sparse-reward environment → sparse reward
- sparsity
- spatial intelligence
- SPD → parameter decomposition
- speaker diarisation → speaker diarization
- speaker diarization
- speaker identification → speaker verification
- speaker recognition → speaker verification
- speaker verification
- Spearman's rank correlation → correlation
- spec → model specification
- spec gaming → specification gaming
- SpecDec → speculative decoding
- spec-driven development
- spec-first development → spec-driven development
- special token
- specialized AI → narrow AI
- specification exploitation → specification gaming
- specification gaming
- specificity
- spectral clustering
- spectral norm → norm
- spectral normalization → normalization layer
- spectrogram
- spectrum → eigenvector
- speculative decoding
- speculative sampling → speculative decoding
- speech activity detection → voice activity detection
- speech cloning → voice cloning
- speech enhancement → source separation
- speech language model → audio language model
- speech recognition
- speech separation → source separation
- speech synthesis → text-to-speech
- speech synthesizer → text-to-speech
- SpeechLM → audio language model
- speech-to-speech → speech-to-speech model
- speech-to-speech model
- speech-to-speech translation → speech-to-speech model
- speech-to-text → speech recognition
- spherical score → proper scoring rule
- spicy autocomplete → fancy autocomplete
- spider → web crawler
- Spider-Man neuron → multimodal neuron
- spike → loss spike
- spiking neural network
- spiking neuron → spiking neural network
- spiritual bliss attractor → spiritual bliss attractor state
- spiritual bliss attractor state
- splat → Gaussian splatting
- SP-LIME → LIME
- split conformal prediction → conformal prediction
- split prefill → chunked prefill
- split testing → A/B testing
- SPM → fill-in-the-middle
- spoken dialogue system → conversational AI
- spoken language model → audio language model
- sponge attack → sponge example
- sponge example
- spurious association → spurious correlation
- spurious correlation
- spurious feature → spurious correlation
- Sputnik moment → ChatGPT moment
- SQuAD
- SQuAD 1.1 → SQuAD
- SQuAD 2.0 → SQuAD
- squared error loss → mean squared error
- squared hinge loss → hinge loss
- squashing function → sigmoid function
- squeeze-and-excitation block → squeeze-and-excitation network
- squeeze-and-excitation network
- squiggle maximizer → paperclip maximizer
- SR → super-resolution
- SR 11-7 → model risk management
- SRAM
- SRGAN → super-resolution
- SRI → SRI International
- SRI Artificial Intelligence Center → SRI International
- SRI International
- s-risk → suffering risk
- SRM → empirical risk minimization
- SSE → streaming
- SSI → Safe Superintelligence Inc.
- SSL → self-supervised learning
- SSM → state space model
- Stability → Stability AI
- Stability AI
- stability–plasticity dilemma
- stability-plasticity tradeoff → stability–plasticity dilemma
- stability–plasticity trade-off → stability–plasticity dilemma
- Stable Diffusion
- Stable Diffusion 3 → Stable Diffusion
- Stable Diffusion 3.5 → Stable Diffusion
- Stable Diffusion XL → Stable Diffusion
- stable model → answer set programming
- stable model semantics → answer set programming
- StableHLO → XLA
- stacked denoising autoencoder → denoising autoencoder
- stacked DRAM → high-bandwidth memory
- stacked ensemble → stacking
- stacked generalization → stacking
- stacked regression → stacking
- stacking
- staged deployment
- staged release → staged deployment
- staged rollout → canary deployment
- staggered release → staged deployment
- Stan → probabilistic programming
- standard deviation → variance
- standard error → confidence interval
- standardization → feature scaling
- Stanford HAI
- Stanford Institute for Human-Centered Artificial Intelligence → Stanford HAI
- Stanford Question Answering Dataset → SQuAD
- Stanford Research Institute → SRI International
- Stanford Research Institute Problem Solver → STRIPS
- STaR → self-taught reasoner
- StarAI → Markov logic network
- Stargate
- Stargate LLC → Stargate
- Stargate Project → Stargate
- Stargate UAE → Stargate
- state
- state AI law moratorium → AI preemption
- state explosion → combinatorial explosion
- state of the art
- state space duality → Mamba
- state space model
- state space search
- state–action trajectory → trajectory
- state–action value function → Q-function
- state–action–reward–state–action → SARSA
- statement logic → propositional logic
- Statement on AI Risk
- Statement on Superintelligence
- state-value function → value function
- static evaluation function → evaluation function
- static graph → computational graph
- static RAM → SRAM
- static random-access memory → SRAM
- static word embedding → word embedding
- stationary distribution → Markov chain
- stationary point → local minimum
- statistical classification → classification
- statistical decision theory → decision theory
- statistical fairness → group fairness
- statistical hypothesis testing → hypothesis testing
- statistical language model → language model
- statistical learning → machine learning
- statistical learning theory
- statistical machine translation → machine translation
- statistical parity → demographic parity
- statistical pattern recognition → pattern recognition
- statistical power → hypothesis testing
- statistical relational learning → Markov logic network
- statistical significance → hypothesis testing
- statistically significant → hypothesis testing
- STE → straight-through estimator
- stealth nerf → nerfed
- steepest ascent → hill climbing
- steepest descent → gradient descent
- steer → steering
- steerability
- steerable → steerability
- steering
- steering vector
- steganographic → steganography
- steganographic collusion → steganography
- steganography
- stem separation → source separation
- stemmer → stemming
- stemming
- step → training step
- step decay → learning rate schedule
- step distillation → diffusion distillation
- step size → learning rate
- step-by-step reasoning → chain of thought
- Stephen M. Omohundro → Steve Omohundro
- Stephen Omohundro → Steve Omohundro
- step-level reward model → process reward model
- step-level supervision → process supervision
- stepwise relative reachability → relative reachability
- stepwise selection → feature selection
- stereo matching → depth estimation
- STFT → spectrogram
- stigmergy → swarm intelligence
- stochastic
- stochastic game → multi-agent reinforcement learning
- stochastic gradient boosting → gradient boosting
- stochastic gradient descent
- stochastic Hopfield network → Boltzmann machine
- stochastic interpolants → flow matching
- stochastic local search → local search
- stochastic neighbor embedding → t-SNE
- stochastic parameter decomposition → parameter decomposition
- stochastic parrot
- stochastic parrots → stochastic parrot
- stochastic process → stochastic
- stochastic variable → random variable
- stochastic variational inference → variational inference
- stochastic weight averaging → exponential moving average of weights
- stochasticity → stochastic
- stop condition
- stop parameter → stop sequence
- stop sequence
- stop string → stop sequence
- stop token → stop sequence
- stop-button problem → shutdown problem
- stopping condition → stop condition
- stopping criterion → stop condition
- straggler
- straggler mitigation → straggler
- stragglers → straggler
- straight-line distance → Euclidean distance
- straight-through estimator
- straight-through gradient → straight-through estimator
- straight-through Gumbel-softmax → Gumbel-softmax
- Strategic Computing → Strategic Computing Initiative
- Strategic Computing Initiative
- Strategic Computing Program → Strategic Computing Initiative
- strategic deception → AI deception
- strategic underperformance → sandbagging
- stratified k-fold → cross-validation
- Strawberry → OpenAI o1
- strawberry problem
- strawberry test → strawberry problem
- stream → streaming
- streamed response → streaming
- streaming
- Streaming DiLoCo → DiLoCo
- streaming inference → streaming
- streaming learning → online learning
- Streamlit → Gradio
- strictly proper scoring rule → proper scoring rule
- stride
- strided convolution → stride
- STRIPS
- STRIPS assumption → STRIPS
- STRIPS representation → STRIPS
- strong AI
- strong artificial intelligence → strong AI
- strong baseline → baseline
- strong emergence → emergence
- strong law of large numbers → law of large numbers
- strong learner → boosting
- strong longtermism → longtermism
- Strong REJECT → StrongREJECT
- Strong, Robust Evaluation of Jailbreaks at Evading Censorship Techniques → StrongREJECT
- strong scaling hypothesis → scaling hypothesis
- StrongREJECT
- StrongREJECT benchmark → StrongREJECT
- structural causal model
- structural credit assignment → credit assignment problem
- structural equation model → structural causal model
- structure from motion
- structured access
- structured data → tabular data
- structured generation → constrained decoding
- structured output → structured outputs
- structured outputs
- structured pruning → pruning
- structured state space model → state space model
- structured state space sequence model → state space model
- STS → semantic textual similarity
- STS Benchmark → semantic textual similarity
- STS-B → semantic textual similarity
- STSb → semantic textual similarity
- STT → speech recognition
- Stuart J. Russell → Stuart Russell
- student model → knowledge distillation
- style cloaking → Glaze
- style transfer
- style-based generator → StyleGAN
- StyleGAN
- StyleGAN2 → StyleGAN
- StyleGAN2-ADA → StyleGAN
- StyleGAN3 → StyleGAN
- subagent
- sub-agent → subagent
- subject-driven generation → DreamBooth
- subjective experience → consciousness
- subject–predicate–object triple → knowledge graph
- subliminal learning
- subsampling → pooling
- subset selection → feature selection
- substantial modification
- substantially modified → substantial modification
- substitute model → surrogate model
- substrate independence
- substrate neutrality → substrate independence
- substrate-independent → substrate independence
- subsumption → subsumption architecture
- subsumption architecture
- subtask decomposition → task decomposition
- subword token → token
- subword tokenization → tokenization
- successionism
- successionist → successionism
- successive approximation → value iteration
- successive halving → Hyperband
- success@k → recall@k
- successor state axiom → situation calculus
- suffering risk
- sufficiency (fairness criterion) → predictive parity
- Mustafa Suleyman
- summarisation → text summarization
- summarization → text summarization
- Summit of the Future → Global Digital Compact
- sum-product algorithm → belief propagation
- Suno
- Suno AI → Suno
- Suno Studio → Suno
- super intelligence order → Executive Order 14434
- Super Intelligence (SI) → superintelligence
- super learner → stacking
- superalignment
- Superalignment team → superalignment
- supercluster → AI supercomputer
- superficial alignment → superficial alignment hypothesis
- superficial alignment hypothesis
- SuperGLUE
- superhuman coder
- superhuman intelligence → superintelligence
- superintelligence
- superintelligence alignment → superalignment
- superintelligence control problem → control problem
- superintelligence statement → Statement on Superintelligence
- superintelligent → superintelligence
- superintelligent AI → artificial superintelligence
- superposition
- superposition hypothesis → superposition
- super-resolution
- supervised classification → classification
- supervised fine-tuning
- supervised finetuning → supervised fine-tuning
- supervised learning
- supervised machine learning → supervised learning
- supervised training → supervised learning
- SuperVision → AlexNet
- supervisor pattern → orchestrator–worker pattern
- support vector classifier → support vector machine
- support vector machine
- support vector regression → support vector machine
- support vectors → support vector machine
- support-vector network → support vector machine
- SURF → scale-invariant feature transform
- surprisal → entropy
- surrogate → surrogate model
- surrogate model
- surrogate split → CART
- surveillance capitalism
- surveillance capitalist → surveillance capitalism
- survival analysis
- sustainable AI → Green AI
- Ilya Sutskever
- Richard Sutton
- Sutton's reward hypothesis → reward hypothesis
- SVC → support vector machine
- SVCCA → centered kernel alignment
- SVD → singular value decomposition
- SVM → support vector machine
- SVR → support vector machine
- SWA → sliding window attention
- SWAG → HellaSwag
- swarm intelligence
- swarm of agents → agent swarm
- swarm robotics
- swarm scaling
- swarming → swarm robotics
- SWE-agent
- SWE-bench
- SWE-bench Lite → SWE-bench
- SWE-bench Multilingual → SWE-bench
- SWE-bench Multimodal → SWE-bench
- SWE-bench Pro → SWE-bench
- SWE-bench Verified → SWE-bench
- sweep → hyperparameter tuning
- SwiGLU
- SwiGLU FFN → SwiGLU
- Swin Transformer → vision transformer
- Swish → SiLU
- swish function → SiLU
- Swish-gated linear unit → SwiGLU
- Swiss cheese model → defense in depth
- Switch Transformer → mixture of experts
- sycophancy
- sycophant → sycophancy
- sycophantic → sycophancy
- Sydney
- Sydney (Bing) → Sydney
- symbol grounding → symbol grounding problem
- symbol grounding problem
- symbolic AI
- symbolic artificial intelligence → symbolic AI
- symbolic regression
- symbolism → symbolic AI
- symmetric tree → CatBoost
- symmetry-aware deep learning → geometric deep learning
- synaptic weight → weights
- synchronous data parallelism → data parallelism
- syntactic parsing
- synthetic consciousness → machine consciousness
- synthetic content → synthetic media
- synthetic content labeling → AI-generated content labeling
- synthetic data
- synthetic dataset → synthetic data
- synthetic media
- synthetic media detection → deepfake detection
- synthetic minority oversampling → SMOTE
- Synthetic Minority Over-sampling Technique → SMOTE
- synthetic voice → text-to-speech
- SynthID
- SynthID Detector → SynthID
- SynthID Text → SynthID
- System 1 → System 1 and System 2
- System 1 and System 2
- System 2 → System 1 and System 2
- System 2 thinking → System 1 and System 2
- system card
- system instructions → system prompt
- system message → system prompt
- System One model → decision model
- system prompt
- system prompt extraction → prompt leaking
- system prompt leakage → prompt leaking
- system role → system prompt
- system throughput → throughput
- systematic generalization → systematicity
- systematicity
- systematicity argument → systematicity
- systematicity challenge → systematicity
- systemic risk → general-purpose AI model with systemic risk
- systemic-risk model → general-purpose AI model with systemic risk
- systems reply → Chinese room
- systolic architecture → systolic array
- systolic array