AI PRISM

这里汇总了 AI 领域的关键术语和分类。使用搜索框快速查找。

1. Classification by Capability Level (AI Capabilities)

This classification is based on the intelligence level and adaptability of AI systems, usually divided into three levels.

Artificial Narrow Intelligence (ANI)

AI that excels only in single or limited tasks, cannot generalize to new areas.

Specific task AI (e.g., voice assistant Siri)Image recognition systemRecommendation algorithm (e.g., Netflix recommendation)Chatbot (e.g., ChatGPT)Autonomous navigation systemNarrow AI

Artificial General Intelligence (AGI)

AI that can handle any intellectual task like a human, but currently still in theoretical stage.

Human-level intelligence simulationMultimodal learning systemAutonomous decision frameworkFlexible adaptation to new tasks

Artificial Superintelligence (ASI)

AI that surpasses humans in all fields, potentially bringing ethical challenges, currently a hypothetical concept.

Surpassing human intelligence systemSelf-improving AIGlobal optimization engineHypothetical advanced intelligent entity

2. Classification by Functional Behavior (AI Functionality)

Based on how AI processes information and learns, reflecting evolution from simple responses to advanced consciousness.

Reactive Machines

AI that responds only based on current input, without historical memory or learning ability.

Chess program (e.g., Deep Blue)Basic game AINo-memory decision system

Limited Memory AI

AI that trains using historical data but does not store long-term memory.

Autonomous driving vehicle (e.g., Tesla Autopilot)Chatbot (e.g., early ChatGPT versions)Prediction model

Theory of Mind AI

AI that understands human emotions, beliefs, and intentions, currently under research.

Emotion recognition systemSocial robotAdvanced dialogue AI

Self-Aware AI

AI with self-awareness and desires, currently a science fiction concept.

Consciousness simulation frameworkAutonomous evolution system

3. Classification by Technical Branches (AI Branches)

Focusing on specific technical fields and methods in AI. Includes core branches and their key concepts.

Machine Learning (ML)

Core technology that uses data to train models for prediction or decision-making.

Supervised Learning (classification, regression)Unsupervised Learning (clustering)Reinforcement Learning (RL)Semi-Supervised LearningTransfer LearningEnsemble Learning (Random Forest, XGBoost)Bayesian MethodsSelf-Supervised Learning (SSL)Fine-tuningInference

Deep Learning (DL)

Technology using multi-layer neural networks to process complex data.

Artificial Neural Network (ANN)Convolutional Neural Network (CNN)Recurrent Neural Network (RNN)Long Short-Term Memory (LSTM)Generative Adversarial Network (GAN)Transformer Models (BERT, GPT)Variational Autoencoders (VAEs)Mixture of Experts (MoE)Low-Rank Adaptation (LoRA)

Natural Language Processing (NLP)

Technology for processing and understanding human language.

TokenizationNamed Entity Recognition (NER)Sentiment AnalysisMachine TranslationQuestion Answering SystemsText GenerationSpeech-to-Text (STT)Text-to-Speech (TTS)BERTGPT

Computer Vision (CV)

Technology for analyzing visual data.

Image ClassificationObject Detection (YOLO)Image SegmentationFacial RecognitionOCRMedical Image Analysis

4. Emerging and Advanced AI Concepts

Rapidly developing new technologies in recent years.

Generative AI

AI that creates new content.

Diffusion Models (Stable Diffusion)Large Language Models (LLMs: GPT, Claude)Image Generation (DALL-E, Midjourney)Video Generation (Runway)Code GenerationHallucination

Agentic AI

AI that acts independently like an agent.

Autonomous AgentsMulti-Agent SystemsTask-Specific AI Agents

Foundation Models

Large models trained on broad data for adaptation to multiple tasks.

Large Versatile ModelsPre-trained Models

5. Ethical and Societal Concepts

Non-technical aspects that cannot be ignored in AI development.

AI Ethics

Principles to ensure AI is fair and responsible.

BiasFairnessTransparencyAccountabilityIntellectual Property IssuesResponsible AI

AI Safety & Alignment

Frameworks to prevent AI from going out of control or causing harm.

Value AlignmentRobustnessExplainable AI (XAI)AI AlignmentGroundedness (扎根性/无幻觉)

7. Popular Terms & Buzzwords

Common AI terms for understanding current trends.

Core Terms

Widely used key AI concepts.

LLMsRAG (Retrieval-Augmented Generation)CoT (Chain of Thought)Prompt EngineeringTokenParametersInferenceWeightsTraining DataVibe CodingCanvas (AI UI)

Advanced Concepts

Terms in emerging trends.

Multimodal AIEdge AIQuantum AIRLHF (Reinforcement Learning with Human Feedback)HallucinationDeep Research (深度调研)