Every industry has its own shorthand. In maritime, we throw around terms like noon report or just in time arrival without a second thought, but they mean very little to someone outside the sector. AI is no different, with practitioners cultivating their own shorthand to talk about technical, business and operational complexities. 

However, as artificial intelligence becomes part of everyday business conversations, both in shipping and across other industries, it is increasingly important that our sector keeps track of what people actually mean when they talk about “hallucinations” or “vibe coding.”  Here is a plain-language guide to some of the most common terms you will hear.

1. Vibe coding: Building software by describing what you want in plain language to an AI, then refining the results instead of writing code line by line yourself.
2. Single source of truth (SSOT): A single, authoritative place where specific data lives, ensuring every system or person relies on the exact same version rather than conflicting copies.
3. Hallucination: When an AI model generates output that sounds entirely confident, but is actually factually wrong or completely made up.
4. Prompt engineering: The art of crafting inputs (prompts) to get clearer, more accurate, or more useful answers from an AI model.
5. LLM (Large Language Model): A broad class of AI systems, such as GPT or Claude, trained on vast amounts of text to understand and produce human-like language.
6. Context window: The total amount of text (measured in tokens) an AI model can process and keep in mind at one time, including previous messages in the conversation.
7. Token: A segment of text, roughly a word or part of a word, that serves as the basic unit an AI model reads. This is important because many AI tools charge by token usage, meaning longer inputs and outputs increase costs.
8. Fine-tuning: Training an existing AI model further on a targeted dataset so it performs better in a specific domain or on a particular task.
9. RAG (Retrieval-Augmented Generation): A method where an AI model looks up relevant external documents or live data before responding, instead of relying solely on what it learned during training.
10. Agentic / AI agent: An AI system capable of taking multi-step actions to achieve a goal, such as searching, clicking buttons, or using digital tools, rather than simply answering questions.
11. Model weights: The internal numerical parameters a trained AI uses to make predictions. In short, they represent what the model has actually learned.
12. Inference: The actual process of running a trained model to generate a response, as opposed to the initial training phase.
13. Guardrails: Safety rules and content filters built into an AI system to keep its answers appropriate, secure, and focused on the intended topic.
14. Multimodal: An AI model designed to work with multiple types of media at once, including text, images, audio, and video.
15. Ground truth: Verified, accurate real-world data used as a benchmark to check whether an AI model is giving correct answers.

Just as understanding nautical terminology helps you follow a discussion on vessel operations, getting comfortable with these AI terms makes it much easier to participate in industry conversations. Whether you are in a leadership meeting discussing new software adoption or attending a panel on the future of shipping, you do not need to be a computer scientist to contribute. You simply need a reliable starting point, and hopefully this guide serves as a helpful reference.

Words: Namrata Nadkarni, assisted by AI. Photo by Sandy Millar on Unsplash

Like this article? Click here to sign up for our free monthly ‘De-Mystifying Digitalisation and AI in Maritime’ newsletter and get the latest cross-sector expert insights, news, interviews, videos and case studies straight to your inbox. You can also buy a ticket to the Demystifying AI in Maritime 2026 London Summit here.