# terminology

[Corpus index](../index.txt)

- [agent](../terminology/agent.txt) — エージェント
- [AI alignment](../terminology/ai_alignment.txt) — AIアラインメント; alignment; アラインメント; アライメント; 整合; AIアライメント
- [AI control](../terminology/ai_control.txt) — AIコントロール; AI control research; AI制御
- [AI development pause](../terminology/ai_development_pause.txt) — AI開発の一時停止; AI development moratorium; AI moratorium; AI pause; pause on AI development; AIモラトリアム
- [AI existential risk](../terminology/ai_existential_risk.txt) — AI x-risk; existential risk from AI
- [AI governance](../terminology/ai_governance.txt) — artificial intelligence governance
- [AI R&D automation](../terminology/ai_r_and_d_automation.txt) — AI 研究開発の自動化; automation of AI R&D; AI 研究開発自動化
- [AI race dynamics](../terminology/ai_race_dynamics.txt) — AI開発競争の力学; AI arms race dynamics; AI race pressure; competitive pressure in AI development; 軍拡競争; 競争圧力
- [AI risk](../terminology/ai_risk.txt) — AIリスク; artificial intelligence risk
- [AI safety](../terminology/ai_safety.txt) — AI安全性; artificial intelligence safety
- [AI Safety Level](../terminology/ai_safety_level.txt) — AI Safety Levels; ASL; ASLs
- [alignment audit](../terminology/alignment_audit.txt) — alignment auditing; alignment audits
- [Alignment by Default](../terminology/alignment_by_default.txt) — アラインメント・バイ・デフォルト; alignment-by-default
- [alignment failure](../terminology/alignment_failure.txt) — アラインメント失敗; AI alignment failure
- [alignment faking](../terminology/alignment_faking.txt) — アラインメント偽装; alignment-faking; fake alignment; alignment fake; 猫を被る
- [alignment problem](../terminology/alignment_problem.txt) — AIアラインメント問題; AI alignment problem; the alignment problem; アラインメント問題
- [alignment risk](../terminology/alignment_risk.txt) — AI alignment risk; misalignment risk
- [alignment target](../terminology/alignment_target.txt) — アラインメントの目標基準; target of alignment; アラインメント対象
- [artificial general intelligence](../terminology/artificial_general_intelligence.txt) — 汎用人工知能; AGI; general AI
- [artificial intelligence](../terminology/artificial_intelligence.txt) — 人工知能; AI
- [artificial superintelligence](../terminology/artificial_superintelligence.txt) — 人工超知能; ASI; superintelligent AI
- [attack policy](../terminology/attack_policy.txt) — attack policies; red-team attack policy
- [automated alignment](../terminology/automated_alignment.txt) — アラインメント研究の自動化; AI aligning AI; AI-assisted alignment; automated AI alignment; automated alignment research; AI支援アラインメント
- [base objective](../terminology/base_objective.txt) — ベース目的関数; base-objective; ベース目的
- [base optimizer](../terminology/base_optimizer.txt) — ベースオプティマイザ; base-optimizer; ベースオプティマイザー
- [Bayesian inference](../terminology/bayesian_inference.txt) — ベイズ推定; Bayesian updating
- [behavioral objective](../terminology/behavioral_objective.txt) — 行動から推定される目的; behavioural objective; 行動上の目的; 行動目的
- [benchmark](../terminology/benchmark.txt) — ベンチマーク; benchmark evaluation
- [capability](../terminology/capability.txt) — 能力; capabilities; AI能力
- [capability elicitation](../terminology/capability_elicitation.txt) — capabilities elicitation; eliciting capabilities
- [capability evaluation](../terminology/capability_evaluation.txt) — 能力評価; capability eval; capability evaluations
- [capability generalization](../terminology/capability_generalization.txt) — 能力の汎化; capabilities generalisation; capabilities generalization; capability generalisation
- [capability threshold](../terminology/capability_threshold.txt) — capability thresholds
- [catastrophic risk](../terminology/catastrophic_risk.txt) — 壊滅的リスク; catastrophic risks; 破局的リスク; 破滅的リスク
- [Coherent Extrapolated Volition](../terminology/coherent_extrapolated_volition.txt) — 首尾一貫した外挿的意志; CEV
- [collusion](../terminology/collusion.txt) — AI collusion; actor-monitor collusion; model collusion; policy-monitor collusion
- [compute governance](../terminology/compute_governance.txt) — AI compute governance
- [control evaluation](../terminology/control_evaluation.txt) — AI control eval; AI control evaluation; control eval; control evaluations
- [control protocol](../terminology/control_protocol.txt) — AI control protocol; control protocols
- [control safety case](../terminology/control_safety_case.txt) — AI control safety case; control safety cases
- [corrigibility](../terminology/corrigibility.txt) — 訂正可能性; corrigible; 修正可能性
- [crunch time](../terminology/crunch_time.txt) — クランチタイム; AI crunch time; AIクランチタイム
- [dangerous capability evaluation](../terminology/dangerous_capability_evaluation.txt) — dangerous capabilities eval; dangerous capabilities evals; dangerous capabilities evaluation; dangerous capabilities evaluations; dangerous capability eval; dangerous capability evals
- [debate](../terminology/debate.txt) — ディベート; AI safety via debate; AI debate
- [deception](../terminology/deception.txt) — 欺瞞; AI deception; strategic deception; デセプション
- [deceptive alignment](../terminology/deceptive_alignment.txt) — 欺瞞的アラインメント; deceptively aligned AI
- [distribution shift](../terminology/distribution_shift.txt) — 分布シフト; distributional shift
- [Eliciting Latent Knowledge](../terminology/eliciting_latent_knowledge.txt) — ELK
- [evaluation](../terminology/evaluation.txt) — 評価; AI eval; AI evals; AI evaluation; AI evaluations; eval; evals; model eval; model evals; model evaluation; model evaluations; AI評価; モデル評価
- [evaluation awareness](../terminology/evaluation_awareness.txt) — 評価認識; eval awareness
- [existential risk](../terminology/existential_risk.txt) — 存亡リスク; existential risks; x-risk; x risk; 実存的リスク; 存亡的リスク
- [factored cognition](../terminology/factored_cognition.txt)
- [first critical try](../terminology/first_critical_try.txt) — 最初の危険な試行; first critical tries; 最初の重要な試行
- [Friendly AI](../terminology/friendly_ai.txt) — フレンドリーAI; FAI; Friendly Artificial Intelligence
- [frontier model](../terminology/frontier_model.txt) — frontier AI model; frontier AI models; frontier models
- [generalization](../terminology/generalization.txt) — 汎化; generalisation
- [goal](../terminology/goal.txt) — 目標
- [goal-directed behavior](../terminology/goal_directed_behavior.txt) — 目標指向的行動; goal-directed; goal directed behavior
- [goal-guarding scheming](../terminology/goal_guarding_scheming.txt) — 目標保護型スキーミング; goal guarding scheming; goal-guarding schemer
- [goal misgeneralization](../terminology/goal_misgeneralization.txt) — 目標の誤汎化; goal mis-generalization; 目標の誤一般化
- [Goodhart's Law](../terminology/goodharts_law.txt) — グッドハートの法則
- [human preferences](../terminology/human_preferences.txt) — 人間の選好
- [human values](../terminology/human_values.txt) — 人間の価値観
- [industrial explosion](../terminology/industrial_explosion.txt) — 産業爆発; AI industrial explosion
- [inner alignment](../terminology/inner_alignment.txt) — インナーアラインメント; inner alignment problem
- [inner misalignment](../terminology/inner_misalignment.txt) — インナー・ミスアラインメント; inner alignment failure
- [instrumental convergence](../terminology/instrumental_convergence.txt) — 道具的収束; convergent instrumental goals
- [instrumental goal](../terminology/instrumental_goal.txt) — 道具的目標; instrumental subgoal; 手段的目標
- [intelligence explosion](../terminology/intelligence_explosion.txt) — 知能爆発; AI intelligence explosion
- [intended goal](../terminology/intended_goal.txt) — 意図された目標; designer's intended goal; programmer's intended goal; 設計者の意図した目標
- [interpretability](../terminology/interpretability.txt) — 解釈可能性; AI interpretability; ML interpretability
- [iterated amplification](../terminology/iterated_amplification.txt) — IDA; iterated distillation and amplification
- [latent knowledge](../terminology/latent_knowledge.txt)
- [loss of control](../terminology/loss_of_control.txt) — 制御喪失; AI loss of control; loss of human control
- [machine learning](../terminology/machine_learning.txt) — 機械学習; ML
- [mechanistic interpretability](../terminology/mechanistic_interpretability.txt) — メカニスティック・インタープリタビリティ; mech interp; mechanistic interp; mechinterp; メカインタープ
- [mesa-objective](../terminology/mesa_objective.txt) — メサ目的関数; mesa objective; メサ目的
- [mesa-optimization](../terminology/mesa_optimization.txt) — メサ最適化; mesa optimization
- [mesa-optimizer](../terminology/mesa_optimizer.txt) — メサオプティマイザ; mesa optimizer
- [misaligned AGI](../terminology/misaligned_agi.txt) — ミスアラインドAGI; misaligned artificial general intelligence
- [misaligned AI](../terminology/misaligned_ai.txt) — ミスアラインドAI; misaligned artificial intelligence; ミスアラインド
- [misaligned power-seeking](../terminology/misaligned_power_seeking.txt) — MAPS; misaligned power seeking
- [misalignment](../terminology/misalignment.txt) — ミスアラインメント; AI misalignment
- [neural network](../terminology/neural_network.txt) — ニューラルネットワーク; neural networks
- [objective](../terminology/objective.txt) — 目的; objective function
- [objective robustness](../terminology/objective_robustness.txt) — 目的の頑健性; 目的頑健性
- [ontology identification problem](../terminology/ontology_identification_problem.txt) — ontology identification
- [optimization](../terminology/optimization.txt) — 最適化
- [optimizer](../terminology/optimizer.txt) — オプティマイザ; 最適化器; オプティマイザー
- [Orthogonality Thesis](../terminology/orthogonality_thesis.txt) — 直交性テーゼ; 直交仮説
- [outcome supervision](../terminology/outcome_supervision.txt) — outcome-based supervision; outcome-based training
- [outer alignment](../terminology/outer_alignment.txt) — アウターアラインメント; outer alignment problem; reward misspecification problem
- [outer misalignment](../terminology/outer_misalignment.txt) — アウター・ミスアラインメント; outer alignment failure; reward misspecification
- [oversight](../terminology/oversight.txt) — 監督; AI oversight; human oversight; オーバーサイト; 監視
- [p(doom)](../terminology/p_doom.txt) — probability of doom; Pdoom; 破滅確率; 破滅の確率
- [paperclip maximizer](../terminology/paperclip_maximizer.txt) — クリップ最大化器; paperclip maximiser
- [pivotal act](../terminology/pivotal_act.txt) — AI-enabled pivotal act; pivotal achievement
- [power concentration](../terminology/power_concentration.txt) — 権力集中; centralization of power; concentration of power; 権力の集中
- [power-seeking](../terminology/power_seeking.txt) — パワーシーキング; power seeking
- [preferences](../terminology/preferences.txt) — 選好
- [process supervision](../terminology/process_supervision.txt) — process-based supervision; process-based training
- [prosaic alignment](../terminology/prosaic_alignment.txt) — プロセイック・アラインメント; prosaic AI alignment
- [proxy objective](../terminology/proxy_objective.txt) — 代理目的; proxy objective function; プロキシ目的; 代理目的関数
- [recursive self-improvement](../terminology/recursive_self_improvement.txt) — 再帰的自己改善; RSI; recursive AI self-improvement
- [red teaming](../terminology/red_teaming.txt) — AI red teaming; red-teaming; red team testing
- [reinforcement learning](../terminology/reinforcement_learning.txt) — 強化学習; RL
- [reinforcement learning from human feedback](../terminology/reinforcement_learning_from_human_feedback.txt) — 人間のフィードバックによる強化学習; RLHF; reinforcement learning with human feedback
- [resource acquisition](../terminology/resource_acquisition.txt) — 資源獲得; resource-seeking; resource seeking
- [Responsible Scaling Policy](../terminology/responsible_scaling_policy.txt) — RSP; RSPs; Responsible Scaling Policies
- [reward](../terminology/reward.txt) — 報酬; reinforcement signal; reward signal
- [reward hacking](../terminology/reward_hacking.txt) — 報酬ハッキング; reward-hacking; リワードハッキング
- [reward tampering](../terminology/reward_tampering.txt) — 報酬改ざん; reward channel tampering; reward-channel tampering; reward process tampering; 報酬の改ざん
- [risk](../terminology/risk.txt) — リスク
- [safety case](../terminology/safety_case.txt) — AI safety case; AI safety cases; safety cases
- [sandbagging](../terminology/sandbagging.txt) — サンドバッギング; AI sandbagging; strategic underperformance
- [scalable oversight](../terminology/scalable_oversight.txt) — スケーラブル・オーバーサイト; スケーラブルオーバーサイト
- [scheming](../terminology/scheming.txt) — スキーミング; AI scheming; schemer
- [security mindset](../terminology/security_mindset.txt) — セキュリティ・マインドセット; security-oriented mindset
- [self-preservation](../terminology/self_preservation.txt) — 自己保存; self preservation
- [sharp left turn](../terminology/sharp_left_turn.txt) — シャープ・レフト・ターン; hard left turn
- [situational awareness](../terminology/situational_awareness.txt) — 状況認識; AI situational awareness
- [specification gaming](../terminology/specification_gaming.txt) — 仕様ゲーミング; reward gaming; specification gaming behavior
- [takeoff speed](../terminology/takeoff_speed.txt) — テイクオフ速度; AI takeoff speed; speed of AI takeoff; 離陸速度
- [takeover-seeking](../terminology/takeover_seeking.txt) — AI takeover-seeking; takeover seeking; 乗っ取り
- [technology explosion](../terminology/technology_explosion.txt) — 技術爆発; AI technology explosion; technological explosion
- [terminal goal](../terminology/terminal_goal.txt) — 最終目標; final goal; ultimate goal; 終端目標
- [training](../terminology/training.txt) — 学習; model training
- [training-gaming](../terminology/training_gaming.txt) — トレーニング・ゲーミング; playing the training game; training game; training gaming; training-gamer; 訓練ゲーム; 訓練ゲーミング
- [training objective](../terminology/training_objective.txt) — 学習目的関数; training objective function; 訓練目的関数
- [trusted monitoring](../terminology/trusted_monitoring.txt) — trusted model monitoring; trusted monitoring protocol
- [untrusted monitoring](../terminology/untrusted_monitoring.txt) — untrusted model monitoring; untrusted monitoring protocol
- [utility function](../terminology/utility_function.txt) — 効用関数; utility functions
- [value fragility](../terminology/value_fragility.txt) — 価値の脆弱性; fragility of value; value is fragile
- [values](../terminology/values.txt) — 価値観
- [weak AGI](../terminology/weak_agi.txt) — 弱AGI; weak artificial general intelligence; 弱いAGI
- [weak-to-strong generalization](../terminology/weak_to_strong_generalization.txt) — W2SG; weak to strong generalization
- [world model](../terminology/world_model.txt) — world-model