Updated: May 31, 2023 (February 20, 2023)
Charts & IllustrationsWhat Is Generative AI?
Generative AI can create, or generate, entirely new, original content based on a simple textual request from a user. The illustration shows a screen capture from a conversational chatbot, called ChatGPT, being asked about itself, and an image generated by DALL-E from the query “three dogs dancing on the Moon during sunset.”
Generative AI technologies from OpenAI, one of the early pioneers in the space, rely on a natural language machine-learning application called Generative Pretrained Transformer (GPT-3.5). Like all machine learning applications, GPT-3.5 relies on training data and training algorithms to construct a model that it uses for runtime inferencing, which for generative AI means interpreting queries and generating a response. Such Large Language Model (LLMs) are trained on enormous amounts of data, which includes much of the content of the World Wide Web, Wikipedia, and other sources. GPT-3.5’s inferencing reputedly relies on some 175 billion parameters (weights).
Generative AI relies on an “alphabet soup” of algorithms, which include the following:
Atlas Members have full access
Get access to this and thousands of other unbiased analyses, roadmaps, decision kits, infographics, reference guides, and more, all included with membership. Comprehensive access to the most in-depth and unbiased expertise for Microsoft enterprise decision-making is waiting.
Membership OptionsAlready have an account? Login Now