About Generative Artificial Intelligence

Generative AI (GenAI) is a general term for artificial intelligence that creates new content by generating new data samples that are similar to the training set. These generative models learn patterns, structures, and features from the input data and can create content with similar characteristics.

GenAI can be used to create text, music, and images that mimic human creation with varying degrees of success. ChatGPT is currently the most well-known GenAI application and is a sophisticated chatbot that has been trained on an enormous collection of text data to develop an understanding of the patterns and structures of human language. ChatGPT was developed by OpenAI.

The model behind ChatGPT, which relies on that vast collection of text, is called a “large language model” (LLM).

LLMs have been trained on digitized books, articles, websites, and other types of text to learn patterns in natural and written language. In response to prompts from users, ChatGPT can generate text that is coherent and convincingly human-like in seconds. It can summarize historical events, write an essay or basic computer code, translate a passage, or even compose poetry and songs. Examples of uses that have already emerged include using it as a research assistant, a proofreader, a brainstorming aid, a calculator, and many more.

However, outputs fall short of demonstrating the higher levels of learning needed to succeed in a rigorous college setting. Additionally, it’s also sometimes wrong, and with great confidence.

GenAI In-Depth: The Science and Capabilities of GenAI

Advances in deep learning techniques, specifically transformer architectures, have enabled more powerful and efficient modeling of complex relationships in the data. Learn more...about the science and capabilities of GenAI

What is ChatGPT?

Here is ChatGPT’s response when asked “What is ChatGPT?”

Response to What is ChatGPT?

While ChatGPT has been the most discussed machine learning tool of late, alternatives are available or in development (e.g. Google Bard, etc.). Tools like Stable Diffusion or DALL-E 2 have been used to generate surprisingly beautiful, detailed, original, and realistic images based on text prompts. New tools are emerging on a daily basis and so it can be hard to keep up. One of the goals of the GenAI site is to keep you up-to-date with reviews and tips regarding the latest models, as well as guidance for responsible use.

Can an AI make an interesting or beautiful image?

Below are two different image GenAI responses to the prompt “Create an image of Ann Arbor in the style of Piet Mondrian.” The first was generated by DALL-E, OpenAI’s image generator, and the second is from Craiyon, which was built on DALL-E mini.

Piet Mondrian was a Dutch abstract artist in the early 20th Century. You can see examples of his original works from the Museum of Modern Art and learn more about him and his art at The Art Story, an online resource for art history education.

DALL-E image of Ann Arbor in the style of Piet Mondrian

Craiyon image of Ann Arbor in the style of Piet Mondrian

GenAI’s ability to mimic effectively depends on the amount of training data it has seen. The more data it's trained on, the more compelling its replications can be. See the example below for an instance where the model has comparatively less training data—we asked DALL-E to “generate an image of a city in the style of Amy Casey.” Casey is a contemporary American artist who paints fantastical representations of detailed small houses, industrial buildings, and more recently, natural and forest scenes.

DALL-E image of a city in the style of Amy Casey

What is ChatGPT?

ChatGPT is an advanced conversational AI language model developed by OpenAI. It is based on the GPT-3.5 architecture, which stands for "Generative Pre-trained Transformer 3.5." It is designed to generate human-like responses based on the input it receives.

GPT-3.5 is trained on a massive amount of text data from diverse sources, such as books, articles, and websites, allowing it to acquire a wide range of knowledge and language patterns. It can understand and generate text in various languages and has been trained to perform various tasks, including answering questions, providing explanations, writing essays, generating code snippets, and much more.

As a language model, ChatGPT can engage in conversation, making it suitable for chatbots, virtual assistants, and other interactive applications. It has the ability to understand context, follow the flow of a conversation, and produce coherent and contextually relevant responses.

It is important to note that as an AI language model, ChatGPT does not possess true understanding or consciousness. It operates based on patterns and statistical probabilities learned during training. While it can provide helpful and informative responses, it should not be considered a definitive source of information, and human judgment is always recommended when interpreting its responses.