About GenAI Generative AI (GenAI) refers to a class of AI systems that generate content (mostly text, audio, imagery and video) given user prompts, background information, or other input data. The release of OpenAI’s ChatGPT in late 2022 marked a step change in terms of the capabilities of GenAI systems to use language. ChatGPT and other GenAI systems built on so-called large language models involve training machine learning models that have billions of internal parameters with vast amounts of data. Unlike other, ‘narrow’ AI systems, these GenAI systems are able to perform an extremely broad range of tasks in terms of answering questions, creating text, imagery, video, audio, computer programs based on users’ instructions, and, if they can access other software and hardware functionality, even carry out complex tasks in virtual and physical environments. GenAI systems use models that generate predictions of the next ‘most appropriate’ words, images or sounds given the prompt (for example, a question or instruction) they have received. These predictions are constructed from patterns detected in enormous amounts of historical data (sometimes refined by additional human feedback in their training), but are not based in any way on factual accuracy. While these systems present their responses confidently, and their conversational interface is suggestive of a chatbot ‘persona’ interacting with the user, they have no actual knowledge or understanding of the world. They frequently ‘confabulate’ or ‘fabricate’ answers that are incorrect, misleading, biased or inappropriate. Commercial GenAI vendors and those using GenAI in the background for some functionality in their IT products will frequently add ‘human-coded’ additional behaviours to avoid some of these problems, but these are unreliable and may serve the vendor’s interests more than the user’s.This notwithstanding, GenAI systems offer non-expert users many functionalities that can be used purposefully, as long as their output is reviewed, critically analysed, and not expected to substitute human creativity, insight or judgment. This article was published on 2024-11-08