AI Awareness

x

Generative AI

Generative artificial intelligence works by taking content around information and creating new data. This can be done by text, images, audio, or video, and all the work can be done by using information that already exists. In a programming language like the LLM (large language model), you can take information and use it to process information like summaries and make information like ChatGPT. The "large" in the LLms represents the language, model are instructed on extensive data. This system works by deciding on what the user wants based on tokens and choosing one of them. "The model is just predicting the next word. It doesn't understand. But to a user experimenting with it, it appears incredibly capable, even though it has significant blind spots." (Monahan 2023). Rayid Ghani, professor of machine learning at Carnegie Mellon University's Heinz College of Information Systems and Public Policy. Different artificial intelligence models are programmed to choose tokens. This works by selecting the most frequent fourth words on the first try. If a person tries it another time, they will find out that the same prompt submitted didn't get the same output at first. This is why, as people, we sometimes ask AI chatbot models that same question and hope to get different answers.

Note: Copilot and ChatGPT use token process to write computer code.

x

Generative AI Example

A perfect example was the text-to-image model like DALL-E and Stable Diffusion. Pictures and the details of the training program. It understands the common traits and the similarity between words and visual elements. After typing a specific prompt for an image, it checks for common phrases to generate the new image based on the description of the image. Consequently, the models make art that nobody has ever known if they have existed. Down below is a picture of Carnegie Mellon University mascot Scotty Dog dancing. An artificial intelligence prompt generated this, but Artists are still debating whether this creative art was based on creative or really plagiarism.