A guide to the language of artificial intelligence

Artificial intelligence is changing the world and inventing a new language to describe it. For tech professionals, the emergence of terms like large language models (LLMs), retrieval-augmented generation (RAG), and reinforcement learning from human feedback (RLHF) can create professional insecurity. To address this, a living glossary of terms serves to demystify the evolving lexicon of the industry.

A central point of discussion is artificial general intelligence (AGI), a term that generally refers to AI that is more capable than the average human at many, if not most, tasks. Defining AGI remains a challenge, with different organizations offering distinct benchmarks. Sam Altman, the CEO of OpenAI, has described AGI as the “equivalent of a median human that you could hire as a co-worker.” Meanwhile, the OpenAI charter defines the concept as highly autonomous systems that outperform humans at most economically valuable work. Google DeepMind, an AI research lab, views AGI as AI that is at least as capable as humans at most cognitive tasks.

Beyond high-level definitions, the industry faces technical hurdles. Chief among these is the phenomenon of hallucinations. Hallucinations produce GenAI outputs that can be misleading and could even lead to real-life risks — with potentially dangerous consequences (think of a health query that returns harmful medical advice). This issue has contributed to a push for specialized models to reduce knowledge gaps.

To understand how these models function, developers look to deep learning, a subset of machine learning inspired by neural networks—algorithmic structures with origins dating back to the 1940s. Deep learning models require millions or more data points to yield good results. To process this data, systems rely on parallelization, which involves performing multiple calculations simultaneously. This is comparable to having 10 employees working on different parts of a project at the same time instead of one employee working sequentially.

These computational methods enable reasoning techniques like chain of thought, where a model breaks a problem into intermediate steps. For example, solving a logic puzzle where a farmer has chickens and cows with 40 heads and 120 legs requires step-by-step calculations to reach the correct answer. By optimizing these processes, the industry aims to make AI systems more reliable and efficient.

Why it matters

This glossary serves as a living reference for tech professionals to navigate the evolving AI landscape, helping to demystify terminology as the field matures.