Asger Risom
The AI terms you’re suddenly expected to understand — explained.
Artificial intelligence has pushed its way into daily life at a remarkable speed. Along with it has come a new vocabulary that often leaves non-experts struggling to follow the conversation.
From boardrooms to politics, AI discussions increasingly rely on technical terms that sound opaque but shape real decisions. Here’s a clear guide to the key words now defining the debate.
Big ideas
AGI, short for artificial general intelligence, refers to a hypothetical form of AI capable of reasoning and problem-solving at a human level.
The singularity describes a speculative moment when AI surpasses human intelligence altogether, potentially reshaping society beyond prediction.
Alignment is a branch of AI safety research focused on ensuring AI systems act in line with human values and intentions.
How AI works
Large language models, or LLMs, are systems trained on massive datasets to generate human-like text. Chatbots such as ChatGPT, Claude, and Gemini are built on this technology.
Transformers are the neural network architecture behind modern LLMs, allowing models to process huge volumes of data at once.
Machine learning refers to systems that improve through data rather than explicit programming, often using neural networks inspired by the human brain.
Power and scale
Compute is the processing power required to train and operate AI systems. Much of it comes from GPUs, or graphics processing units, specialised chips widely supplied by Nvidia.
Data centres are large facilities housing thousands of these chips. Their energy use is often measured in gigawatts, a unit that reflects the enormous electricity demands of AI.
Risks and limits
Bias occurs when AI systems reproduce distortions present in their training data, leading to unfair or inaccurate outcomes.
Hallucinations describe cases where AI confidently produces incorrect information.
Deepfakes are AI-generated images, videos, or audio designed to convincingly imitate real people.
Society and control
Agentic AI refers to systems that can act autonomously without direct prompts.
Prompt engineering involves designing inputs that guide AI toward better outputs.
Open-source AI allows anyone to inspect and modify a model’s code, raising debates over transparency and safety.
Universal basic income is a policy proposal gaining attention amid fears that AI could displace large numbers of jobs.
Understanding these terms has become essential as AI reshapes work, politics, and everyday services.
Source: OpenAi, Business Insider, Nvidia, Meta
