A generative AI model “hallucinates” when it delivers false or misleading information. This happens because of how the models are trained, on massive amounts of data, like articles, books, code and social media posts. In creating its response to a prompt, the AI makes inferences about what words should come next, but it lacks contextual information so it may generate wording that seems plausible but isn’t true. Hallucinations can also result from improper training and/or biased or insufficient data, which leave the model unprepared to answer certain questions.

An LLM is a large language model – basically, a vast trove of data from written material that serves as the foundation for a generative AI tool like ChatGPT or Gemini. When you give an AI chatbot a prompt, it blazes through the LLM data looking for word patterns that match the topic. As it strings together words to craft a response, it’s constantly making predictions about what the next word in the sequence should be. For instance, in the sentence “The ingredients for chili include…” it might suggest “ground beef” or “beans” as the next word in the sequence. It doesn’t understand the words, just the statistical relationships among them.

AI agents – sometimes referred to as agentive AI – are the hot buzzword in generative AI, the anticipated next wave of AI advances for our phones and other devices. These would go beyond chatbots, which respond to your queries or prompts one at a time, to become bona fide AI assistants, like a supercharged Siri or Alexa. They would actively anticipate your needs and be able to handle complex requests, such as travel planning, by working across apps, accounts, data and web searches.

Policies and restrictions placed on AI models to ensure data is handled responsibly and that the model doesn’t create disturbing content.

An aspect of AI that allows computers to learn and make better predictive outcomes without explicit programming. Can be coupled with training sets to generate new content.

Numerical values that give LLMs structure and behavior, enabling them to make predictions.

The datasets, often quite vast, used to help AI models learn, including text, images, code or data.