RD 115 Evans: Definitions

Broadly speaking, the term is used to describe computer systems that can absorb information, process it, and respond in ways similar to humans…. [with machine learning, the] more data the program analyzes, the more it learns and the better it gets at completing a task.” 

The tasks AI can be trained to complete range widely, including recommending a new TV series to you based on your viewing history, driving a car, or evaluating a medical x-ray to determine whether your bone is broken. 

Generative AI is a subset of AI that can learn to create entirely new images, audio, or text using vast amounts of training data. (Wikipedia)

Examples of generative AI programs that have been in the news include OpenAI’s ChatGPT, which creates text in response to questions and prompts, and DALL-E, which creates new images that correspond to a text-based prompt.

Generative AI tools are based on Large Language Models, where billions of phrases, words, and texts gathered from the web and queried over and over in order to learn how to construct human-sounding answers to questions.  Analytics train the tools to predictively complete sentences. It looks believable, and might get most of the facts right, but it can repeat misinformation or use bias.   It gathers patterns, but cannot use inductive reasoning. Neither does it use inference (reading between the lines), create insights, or use intuition.