U
27 words and 119 other names, in italics, pointing to their entries.
- UAP → universal adversarial perturbation
- UAS → drone
- UAV → drone
- UBI → universal basic income
- UCB → upper confidence bound
- UCB1 → upper confidence bound
- UCF101 → action recognition
- UCIe → chiplet
- UCT → Monte Carlo tree search
- UDA → domain adaptation
- Udio → music generation
- UDT → updateless decision theory
- UI agent → GUI agent
- UKF → Kalman filter
- ultimate goal → terminal goal
- ultraintelligence → ultraintelligent machine
- ultraintelligent machine
- ultra-intelligent machine → ultraintelligent machine
- Ulysses → context parallelism
- UMAP
- umap-learn → UMAP
- UN AI scientific panel → Independent International Scientific Panel on AI
- UN Global Dialogue on AI Governance → Global Dialogue on AI Governance
- UN High-level Advisory Body on AI → High-level Advisory Body on Artificial Intelligence
- UN Scientific Panel on AI → Independent International Scientific Panel on AI
- unacceptable risk → prohibited AI practices
- unacceptable-risk AI → prohibited AI practices
- Unbiased AI Principles → woke AI
- unbounded consumption → sponge example
- uncanny valley
- uncanny valley effect → uncanny valley
- uncertainty quantification
- uncertainty sampling → active learning
- unCLIP → DALL·E
- unconditional basic income → universal basic income
- undercomplete autoencoder → autoencoder
- under-elicitation → capability elicitation
- underfit → underfitting
- underfitted → underfitting
- underfitting
- under-fitting → underfitting
- underflow → numerical stability
- under-refusal → refusal
- under-resourced language → low-resource language
- undersampling → class imbalance
- understanding (AI) → machine understanding
- underthinking → overthinking
- under-trained token → glitch token
- undirected graphical model → Markov random field
- undressing app → nudification app
- unembedding
- un-embedding → unembedding
- unembedding matrix → unembedding
- UNESCO AI ethics recommendation → UNESCO Recommendation on the Ethics of AI
- UNESCO Recommendation on the Ethics of AI
- U-Net
- UNet → U-Net
- unfaithful chain of thought → chain-of-thought faithfulness
- unfaithful reasoning → chain-of-thought faithfulness
- unhobble → unhobbling
- unhobbling
- unhobbling gains → unhobbling
- unidirectional language model → causal language model
- unification
- unified theory of cognition → cognitive architecture
- unifier → unification
- uniform convergence → statistical learning theory
- uniform manifold approximation and projection → UMAP
- uniform soup → model soup
- uniform-cost search → Dijkstra's algorithm
- unify → unification
- unigram language model tokenization → SentencePiece
- unigram tokenizer → SentencePiece
- Unimate → industrial robot
- unique name assumption → closed-world assumption
- unit → artificial neuron
- universal adversarial perturbation
- universal adversarial suffix
- universal approximation → universal approximation theorem
- universal approximation theorem
- universal approximator → universal approximation theorem
- universal basic income
- universal distribution → Solomonoff induction
- universal induction → Solomonoff induction
- universal intelligence
- universal intelligence measure → universal intelligence
- universal jailbreak → jailbreak
- Universal Paperclips
- universal perturbation → universal adversarial perturbation
- universal prior → Solomonoff induction
- universal Turing machine → Turing machine
- universal value function approximator → goal-conditioned reinforcement learning
- universality → universality hypothesis
- universality hypothesis
- unlabeled data → label
- unlearning → machine unlearning
- unmanned aerial vehicle → drone
- unmanned aircraft system → drone
- unrepresentative sample → sampling bias
- unscented Kalman filter → Kalman filter
- unstable training → training instability
- unstructured pruning → pruning
- unsupervised domain adaptation → domain adaptation
- unsupervised learning
- unsupervised machine learning → unsupervised learning
- untrusted monitor → untrusted monitoring
- untrusted monitoring
- up-convolution → transposed convolution
- upconvolution → transposed convolution
- upcycled MoE → sparse upcycling
- upcycling → sparse upcycling
- update step → training step
- updateless decision theory
- updatelessness → updateless decision theory
- UPGMA → hierarchical clustering
- uplift → uplift study
- uplift model → uplift modeling
- uplift modeling
- uplift modelling → uplift modeling
- uplift study
- uplift trial → uplift study
- upload → mind uploading
- uploading → mind uploading
- upper confidence bound
- upper ontology → ontology
- upscaling → super-resolution
- uptraining → grouped-query attention
- UQ → uncertainty quantification
- urn of inventions → vulnerable world hypothesis
- US AI Action Plan → America's AI Action Plan
- US AI Safety Institute → Center for AI Standards and Innovation
- usage policies → usage policy
- usage policy
- usage tier → rate limit
- USAISI → Center for AI Standards and Innovation
- use policy → usage policy
- user prompt → prompt
- user-based collaborative filtering → collaborative filtering
- U-shaped network → U-Net
- U-shaped scaling → inverse scaling
- Jakob Uszkoreit
- utility → utility function
- utility function
- utility indifference
- utility maximization → utility function
- UVFA → goal-conditioned reinforcement learning