FOR EDUCATIONAL AND KNOWLEDGE SHARING PURPOSES ONLY. NOT-FOR-PROFIT. SEE COPYRIGHT DISCLAIMER.

Standard Al Terminology

  • AGl: Al that can think like humans.
  • CoT (Chain of Thought): Al thinking step-by-step.
  • Al Agents: Autonomous programs that make decisions.
  • Al Wrapper: Simplifies interaction with Al models.
  • Al Alignment: Ensuring Al follows human values.
  • Fine-tuning: Improving Al with specific training data.
  • Hallucination: When Al generates false information.
  • Al Model: A trained system for a task.
  • Chatbot: Al that simulates human conversation.
  • Compute: Processing power for Al models.
  • Computer Vision: Al that understands images and videos.
  • Context: Information Al retains for better responses.
  • Deep Learning: Al learning through layered neural networks.
  • dLLM. Diffusion Large Language Model
  • Embedding: Numeric representation of words for Al.
  • Explainability: How Al decisions are understood.
  • Foundation Model: Large Al model adaptable to tasks.
  • Generative Al: Al that creates text, images, etc.
  • GPU: Hardware for fast Al processing.
  • Ground Truth: Verified data Al learns from.
  • Inference: Al making predictions on new data.
  • LLM (Large Language Model): Al trained on vast text data.
  • Machine Learning: Al improving from data experience.
  • MCP (Model Context Protocol): Standard for Al external data access.
  • NLP (Natural Language Processing): Al understanding human language.
  • Neural Network: Al model inspired by the brain.
  • PauseAI. International movement to pause unsafe AGI development until Safe AI can be engineered and delivered with mathematical certainty.
  • Parameters: Al’s internal variables for learning.
  • P(doom): Probability from 0 to 100% that AGI will result in catastrophic outcome for humans.
  • Prompt Engineering: Crafting inputs to guide Al output.
  • Reasoning Model: Al that follows logical thinking.
  • Reinforcement Learning: Al learning from rewards and penalties.
  • RAG (Retrieval-Augmented Generation): Al combining search with responses.
  • Supervised Learning: Al trained on labeled data.
  • TPU: Google’s Al-specialized processor.
  • Tokenization: Breaking text into smaller parts.
  • Training: Teaching Al by adjusting its parameters.
  • Transformer: Al architecture for language processing.
  • Unsupervised Learning: Al finding patterns in unlabeled data.
  • Vibe Coding: Al-assisted coding via natural language prompts.
  • Weights: Values that shape Al learning.
  • X-Risk: Human extinction; the existential risk of Machine intelligence (AI) to replace Homo sapiens (Humans)

FOR EDUCATIONAL AND KNOWLEDGE SHARING PURPOSES ONLY. NOT-FOR-PROFIT. SEE COPYRIGHT DISCLAIMER.