AI Definitions

Understanding the language

Artificial intelligence comes with a rapidly expanding vocabulary. These short definitions provide plain-English explanations of the concepts we use throughout Perspectives.

The aim is not to settle every technical debate, but to establish enough shared language to follow the discussion.

Artificial Intelligence (AI): A broad term for computer systems designed to perform tasks that normally require aspects of human intelligence, such as recognizing patterns, generating language, making predictions, solving problems or taking actions. AI includes many different technologies and levels of capability.
Model: A model is the system produced by training an AI on data so that it can recognize patterns, generate outputs or make predictions. Different models are designed for different purposes, capabilities and levels of complexity.
Training: Training is the process used to develop an AI model by exposing it to large amounts of data and adjusting the system so it becomes better at producing useful outputs. Training is computationally intensive and usually happens before the model is used for inference.
Inference: Inference is what happens when a trained AI model is actually used. When you ask a question, generate an image or give an AI system a task, the model is performing inference to produce its response.
Compute: Compute refers to the processing power required to train and operate AI systems. It depends on hardware such as advanced chips, data centers and the supporting infrastructure needed to run them.
Generative AI: Generative AI refers to systems that can create new content, such as text, images, audio, video or software code, based on patterns learned during training. Large language models are one important form of generative AI.
Large Language Model (LLM): A large language model is an AI model trained on very large amounts of text and other data to understand and generate language. LLMs can answer questions, summarize, write, reason, translate and perform many other language-based tasks.
Agentic AI: Agentic AI refers to systems that can pursue an objective through a sequence of actions rather than simply produce a single response. An agent may plan, use tools, gather information, make decisions and adapt its actions as it works toward a goal.
Artificial General Intelligence (AGI): AGI generally refers to AI with broad, human-level or greater capability across many different cognitive tasks rather than expertise in only one narrow area. There is no universally agreed definition or threshold for when AGI has been reached.
Artificial Superintelligence (ASI): ASI refers to a hypothetical form of AI whose capabilities substantially exceed the best human abilities across a wide range of cognitive domains. ASI has not been demonstrated, and there is significant uncertainty about whether, when or how it might emerge.
Recursive Self-Improvement: Recursive self-improvement refers to the idea that an advanced AI system could contribute to improving the technologies or processes that make future AI systems more capable. If those improvements then help produce still better systems, a reinforcing cycle of improvement could develop.
Alignment: Alignment is the challenge of ensuring that AI systems behave in ways that are consistent with intended human goals, values and constraints. The difficulty is not only defining those goals clearly, but also ensuring that increasingly capable systems interpret and pursue them as intended.
Autonomy: Autonomy refers to the degree to which an AI system can act, make decisions or pursue objectives without continuous human direction. Greater autonomy can increase usefulness, but it can also increase the importance of oversight, safeguards and clearly defined objectives.
Robotics: Robotics involves machines that can sense, move and act in the physical world. When combined with AI, robots can become more capable of understanding their environment, adapting to changing conditions and performing tasks with less direct human control.

Back to Perspectives