There are different types of generative AI tools, including text, image, and sound. Boston College provides students, faculty and staff with access to Microsoft Copilot and Google Gemini in a protected environment.
Generative AI text models attempt to create human-like text based on large amounts of sample training data, often taken from the internet. These systems mimic human communication (without any actual intelligence) by acting like highly complex predictive models of what words would appear, given millions upon millions of example texts. Generative texts have been used to create chatbots, such as the famous ChatGPT. Beyond easing writing, one of its most useful applications has been in helping create/fix programming code. Many have used it to help them brainstorm and draft ideas. It should be noted that these models do not “understand” what they are saying, and are merely repeating patterns of associations found across the millions upon millions of documents powering their output.
Getting Started: Microsoft Copilot, Google Gemini, ChatGPT, Perplexity.ai, HuggingChat, Grammarly
Though not yet as advanced as their image or text equivalents, there are now AI generative sound models. Like other generative AI methods, sound models depend on large training data sets. At the time of this writing, many of these have more limited training data sets associated with them, and are not as flexible as the text and image generators. However, much is changing quickly.
Much like text, AI has famously been applied to the realm of image creation. Models are exposed to billions of images downloaded from the internet and labeled by humans. Then new synthetic images are created via text input based upon the patterns found in those massive data sets. Keep in mind that ethical considerations remain important when working with AI-created images, as they can sometimes raise questions about authenticity and copyright. It's advisable to be aware of these issues and use AI-generated images responsibly in your work.
Getting Started: Dall-E, Stable Diffusion, Midjourney, DreamStudio
AI can be used in combination or in unusual ways. A large number of people use generative text models, like ChatGPT, to help them fix or generate computer code with a simple text input. GitHub, a large online platform for sharing code, has introduced Co-Pilot, an AI feature specifically designed to help write computer code quickly. If you are a programmer and want to use machine learning in your applications, check out the HuggingFace website for resources. In addition, by combining AI text & sound generation, generative AI can also be applied to the realm of video content creation. Applications like Gen-2 can produce video with text, image, or video clip input. Even image generators can be used to produce individual video frames.
Getting Started: GitHub CoPilot, Gen-2
