I
122 words and 347 other names, in italics, pointing to their entries.
- IA → intelligence augmentation
- IAA → inter-annotator agreement
- IAEA for AI
- Ian J. Goodfellow → Ian Goodfellow
- IASR → International AI Safety Report
- IBM → IBM Research
- IBM Research
- IBM Watson
- ICA → independent component analysis
- ICCV
- ICE → individual conditional expectation
- ICE curve → individual conditional expectation
- ICE plot → individual conditional expectation
- ICL → in-context learning
- ICLR
- ICML
- ICOT → Fifth Generation Computer Systems project
- ICRA
- ICV → in-context vector
- ID3
- IDDFS → iterative deepening search
- identity shortcut → residual connection
- IDF → TF-IDF
- idiot's Bayes → naive Bayes
- IDP → document understanding
- IE → information extraction
- IEEE 754 → floating-point format
- IEEE International Conference on Robotics and Automation → ICRA
- IEEE/CVF Conference on Computer Vision and Pattern Recognition → CVPR
- IFEval
- iForest → isolation forest
- IFT → instruction tuning
- if-then commitment
- IHL → international humanitarian law
- IIA → causal abstraction
- i.i.d.
- IID → i.i.d.
- iid → i.i.d.
- IIT → integrated information theory
- IJCAI
- IJCAI Organization → IJCAI
- I-JEPA → joint-embedding predictive architecture
- IK → inverse kinematics
- IK solver → inverse kinematics
- ill-conditioned → numerical stability
- illumination algorithm → quality diversity
- illusion of interpretability → interpretability illusion
- illusionism
- illusionism about consciousness → illusionism
- illusionist → illusionism
- iLQR → optimal control
- ILSVRC → ImageNet
- image caption generation → image captioning
- image captioning
- image classification
- image classifier → image classification
- image completion → inpainting
- image generation model → text-to-image model
- image generator → text-to-image model
- image inpainting → inpainting
- image question answering → visual question answering
- image recognition → image classification
- image segmentation
- image super-resolution → super-resolution
- image tokenizer
- image tokens → image tokenizer
- image translation → image-to-image translation
- image-based sexual abuse → non-consensual intimate imagery
- Imagen
- Imagen 2 → Imagen
- Imagen 3 → Imagen
- Imagen 4 → Imagen
- Imagen Video → Imagen
- ImageNet
- ImageNet challenge → ImageNet
- ImageNet Large Scale Visual Recognition Challenge → ImageNet
- ImageNet-1k → ImageNet
- image-to-image translation
- image-to-video → text-to-video model
- imbalanced classification → class imbalance
- imbalanced data → class imbalance
- img2img → image-to-image translation
- imitation game → Turing test
- imitation learning
- immersion cooling → liquid cooling
- IMO → International Mathematical Olympiad
- IMO gold → IMO Grand Challenge
- IMO gold medal → International Mathematical Olympiad
- IMO Grand Challenge
- impact measure
- impact penalty → impact measure
- impact regularizer → impact measure
- impedance control
- implementational level → Marr's levels of analysis
- implicit bias of gradient descent → implicit regularization
- implicit chain of thought → latent reasoning
- implicit feedback
- implicit gradient regularization → implicit regularization
- implicit neural representation
- implicit regularization
- implicit signals → implicit feedback
- importance ratio → importance sampling
- importance sampling
- importance weight → importance sampling
- importance weights → importance sampling
- impossibility of fairness → fairness impossibility theorem
- impossibility theorem of fairness → fairness impossibility theorem
- improved DDPM → denoising diffusion probabilistic model
- imputation
- impute → imputation
- In re OpenAI Copyright Infringement Litigation → New York Times v. OpenAI
- Inaugurating the Era of Super Intelligence → Executive Order 14434
- Inception
- Inception module → Inception
- Inception network → Inception
- Inception score
- Inceptionism → DeepDream
- Inception-ResNet → Inception
- Inception-v3 → Inception
- Inception-v4 → Inception
- incompleteness theorem → Gödel's incompleteness theorems
- incompleteness theorems → Gödel's incompleteness theorems
- in-context examples → few-shot prompting
- in-context learner → in-context learning
- in-context learning
- in-context memory → working memory
- in-context scheming → scheming
- in-context vector
- in-context vectors → in-context vector
- incorrigibility → corrigibility
- incremental deployment → staged deployment
- incremental response modeling → uplift modeling
- incremental SfM → structure from motion
- independence Bayes → naive Bayes
- independence (fairness criterion) → demographic parity
- independence, separation and sufficiency → group fairness
- independent AI evaluation → third-party AI evaluation
- independent and identically distributed → i.i.d.
- independent component analysis
- Independent International Scientific Panel on AI
- Independent International Scientific Panel on Artificial Intelligence → Independent International Scientific Panel on AI
- independently and identically distributed → i.i.d.
- indicator properties → indicator properties of consciousness
- indicator properties of consciousness
- indifference methods → utility indifference
- indirect discrimination → disparate impact
- indirect injection → indirect prompt injection
- indirect object identification
- indirect prompt injection
- individual conditional expectation
- individual explanation → local explanation
- individual fairness
- induction circuit → induction head
- induction head
- induction heads → induction head
- induction (reasoning) → inductive reasoning
- inductive bias
- inductive conformal prediction → conformal prediction
- inductive inference → inductive reasoning
- inductive logic programming
- inductive program synthesis → program synthesis
- inductive reasoning
- industrial manipulator → industrial robot
- industrial robot
- inertial odometry → odometry
- infer → inference
- inference
- inference API → model API
- inference cloud → inference provider
- inference cost
- inference engine
- inference overhang → compute overhang
- inference pipeline → machine learning pipeline
- inference platform → inference provider
- inference provider
- inference runtime → inference engine
- inference scaling → test-time compute
- inference server → model serving
- inference serving → model serving
- inference to the best explanation → abductive reasoning
- inference-optimal training → overtraining
- inference-time attack → evasion attack
- inference-time compute → test-time compute
- inference-time intervention
- inference-time scaling → test-time compute
- inferencing → inference
- infilling → fill-in-the-middle
- InfiniBand
- infinite-width limit → neural tangent kernel
- Inflection → Inflection AI
- Inflection AI
- in-flight batching → continuous batching
- influence curve → influence function
- influence estimation → data attribution
- influence function
- influence functions → influence function
- influence operation
- influence score → influence function
- infohazard
- Infomax ICA → independent component analysis
- InfoNCE → contrastive learning
- information analysis exception → text and data mining exception
- information bottleneck
- information bottleneck method → information bottleneck
- information bottleneck principle → information bottleneck
- information cocoon → filter bubble
- information disorder → misinformation
- information entropy → entropy
- information extraction
- information gain
- information hazard → infohazard
- information privacy → privacy
- information privacy law → data protection
- information radius → Jensen–Shannon divergence
- information retrieval
- information theory
- information-theoretic → information theory
- informative prior → prior
- informed search → best-first search
- ingestion pipeline → data pipeline
- in-hand manipulation → dexterous manipulation
- inherently interpretable model → interpretable model
- initialization → weight initialization
- injected thought → concept injection
- INLP → iterative nullspace projection
- in-memory computing → analog in-memory computing
- inner alignment
- inner alignment problem → inner alignment
- inner misalignment → inner alignment
- inner optimizer → mesa-optimization
- inner product → dot product
- inoculation prompt → inoculation prompting
- inoculation prompting
- inpaint → inpainting
- inpainting
- input attribution → feature attribution
- input classifier → safety classifier
- input distribution shift → covariate shift
- input drift → data drift
- input prompt → prompt
- input variable → feature
- INR → implicit neural representation
- instability → training instability
- InstaFlow → rectified flow
- instance → example
- instance mask → instance segmentation
- instance normalization → normalization layer
- instance segmentation
- instance-based learning → k-nearest neighbors
- instance-level explanation → local explanation
- InstanceNorm → normalization layer
- Instant-NGP → neural radiance field
- instruct model → instruction-tuned model
- InstructGPT
- instruction adherence → instruction following
- instruction data → instruction dataset
- instruction dataset
- instruction fine-tuning → instruction tuning
- instruction finetuning → instruction tuning
- instruction following
- instruction hierarchy
- instruction privilege → instruction hierarchy
- instruction tuning
- instruction-based image editing
- instruction-following data → instruction dataset
- Instruction-Following Eval → IFEval
- Instruction-Following Evaluation → IFEval
- instruction-following model → instruction-tuned model
- instruction-guided image editing → instruction-based image editing
- instruction-tuned LLM → instruction-tuned model
- instruction-tuned model
- instruction-tuning dataset → instruction dataset
- InstructPix2Pix → instruction-based image editing
- instrumental convergence
- instrumental convergence thesis → instrumental convergence
- instrumental goal
- instrumental subgoal → instrumental goal
- instrumental value → instrumental goal
- INT4 → INT8
- INT8
- int8 inference → LLM.int8()
- integer linear programming → integer programming
- integer program → integer programming
- integer programming
- integer quantization → INT8
- integrated gradients
- integrated information → integrated information theory
- integrated information theory
- intelligence
- intelligence amplification → intelligence augmentation
- intelligence augmentation
- intelligence curse
- intelligence explosion
- intelligent → intelligence
- intelligent agent
- intelligent automation → robotic process automation
- intelligent character recognition → optical character recognition
- intelligent document processing → document understanding
- intelligent machinery → machine intelligence
- intelligent personal assistant → virtual assistant
- intended purpose
- intent alignment
- intent classification → text classification
- intent detection → natural language understanding
- intent-aligned AI → intent alignment
- intentional content → intentionality
- intentional stance
- intentional system → intentional stance
- intentionality
- interactive segmentation → image segmentation
- interactive theorem prover → interactive theorem proving
- interactive theorem proving
- inter-annotator agreement
- intercept → bias
- interchange intervention → activation patching
- interchange intervention accuracy → causal abstraction
- inter-coder agreement → inter-annotator agreement
- interconnect
- interest point → keypoint detection
- interest point detection → keypoint detection
- Interim Measures for the Management of Generative AI Services
- Interim Measures for the Management of Generative Artificial Intelligence Services → Interim Measures for the Management of Generative AI Services
- inter-layer model parallelism → pipeline parallelism
- interleaved reasoning → interleaved thinking
- interleaved thinking
- interleaving → A/B testing
- intermediate checkpoint → checkpoint
- internal covariate shift → batch normalization
- internal deployment
- internal deployment risk → internal deployment
- internal knowledge → parametric knowledge
- internal model → world model
- internal use → internal deployment
- International Agency for AI → IAEA for AI
- international AI agency → IAEA for AI
- International AI Safety Report
- International Conference on Computer Vision → ICCV
- International Conference on Learning Representations → ICLR
- International Conference on Machine Learning → ICML
- international humanitarian law
- International Joint Conference on Artificial Intelligence → IJCAI
- International Mathematical Olympiad
- International Mathematical Olympiad challenge → IMO Grand Challenge
- International Network for Advanced AI Measurement, Evaluation and Science → AI Safety Institute
- International Network of AI Safety Institutes → AI Safety Institute
- International Planning Competition → automated planning
- International Scientific Report on the Safety of Advanced AI → International AI Safety Report
- interpolation threshold → double descent
- interpretability
- interpretability agent → automated interpretability
- interpretability illusion
- interpretability illusions → interpretability illusion
- interpretable → interpretability
- interpretable by design → interpretable model
- interpretable feature → feature (interpretability)
- interpretable model
- InterpretML → explainable boosting machine
- inter-rater agreement → inter-annotator agreement
- inter-rater reliability → inter-annotator agreement
- interruptibility → safe interruptibility
- intersection over union
- inter-token latency → time per output token
- intimate deepfake → non-consensual intimate imagery
- intimate image abuse → non-consensual intimate imagery
- intra-attention → self-attention
- intra-layer model parallelism → tensor parallelism
- intrinsic dimensionality → dimensionality
- intrinsic goal → terminal goal
- intrinsic motivation
- intrinsic reward → intrinsic motivation
- intrinsically interpretable model → interpretable model
- intrinsically motivated reinforcement learning → intrinsic motivation
- introspection
- introspective → introspection
- invariance
- invariant → invariance
- invariant representation → invariance
- inverse document frequency → TF-IDF
- inverse kinematic solver → inverse kinematics
- inverse kinematics
- inverse optimal control → inverse reinforcement learning
- inverse probability → Bayes' theorem
- inverse reinforcement learning
- inverse reward design
- inverse scaling
- inverse scaling law → inverse scaling
- Inverse Scaling Prize → inverse scaling
- inverted bottleneck → MobileNet
- inverted dropout → dropout
- inverted file → inverted index
- inverted index
- inverted residual block → MobileNet
- inverted-U scaling → inverse scaling
- invertible neural network → normalizing flow
- invisible labor (AI) → ghost work
- invisible watermark → AI watermarking
- IOI → indirect object identification
- IOI circuit → indirect object identification
- IOI task → indirect object identification
- IoU → intersection over union
- IP → integer programming
- IP-Adapter → ControlNet
- iPhone moment → ChatGPT moment
- IPI → indirect prompt injection
- IPTO → DARPA
- .ipynb → Jupyter notebook
- IPython Notebook → Jupyter notebook
- IR → information retrieval
- IRD → inverse reward design
- IRL → inverse reinforcement learning
- IROS → ICRA
- irreducible error → irreducible loss
- irreducible loss
- IRT → item response theory
- Irving John Good → I. J. Good
- IS → Inception score
- is all you need → X is all you need
- … is all you need → X is all you need
- is there a seahorse emoji → seahorse emoji
- Isabelle
- Isabelle/HOL → Isabelle
- Isadore Jacob Gudak → I. J. Good
- ISO 42001 → ISO/IEC 42001
- iso-FLOP → IsoFLOP analysis
- IsoFLOP analysis
- IsoFLOP curve → IsoFLOP analysis
- IsoFLOP profile → IsoFLOP analysis
- IsoFLOPs → IsoFLOP analysis
- ISO/IEC 22989
- ISO/IEC 22989:2022 → ISO/IEC 22989
- ISO/IEC 23894
- ISO/IEC 23894:2023 → ISO/IEC 23894
- ISO/IEC 42001
- ISO/IEC 42001:2023 → ISO/IEC 42001
- ISO/IEC JTC 1/SC 42
- ISO/IEC SC 42 → ISO/IEC JTC 1/SC 42
- isolation forest
- isolation-based anomaly detection → isolation forest
- isotonic regression → calibration
- IT model → instruction-tuned model
- Italian brainrot
- item response theory
- item-based collaborative filtering → collaborative filtering
- item-to-item collaborative filtering → collaborative filtering
- iterate averaging → exponential moving average of weights
- iterated amplification
- iterated distillation and amplification → iterated amplification
- iteration → training step
- iteration limit → stop condition
- iteration-level scheduling → continuous batching
- iterative deepening → iterative deepening search
- iterative deepening A* → iterative deepening search
- iterative deepening depth-first search → iterative deepening search
- iterative deepening search
- Iterative Dichotomiser 3 → ID3
- Iterative Dichotomizer 3 → ID3
- iterative FGSM → projected gradient descent attack
- iterative LQR → optimal control
- iterative nullspace projection
- iterative null-space projection → iterative nullspace projection
- ITI → inference-time intervention
- ITL → time per output token
- ITP → interactive theorem proving
- it's over / we're back → it's over, we're so back
- it's over, we're so back
- it's so over → it's over, we're so back
- Alexey Ivakhnenko
- IVFPQ → product quantization
- iWARP → RDMA