Generative Artificial Intelligence

This guide provides information and resources to help learn about generative artificial intelligence (GEN AI or GAI). 

Mural by Beastman (Bradley Eastman) in Christchurch, New Zealand. Photo by Bernard Spragg (CC0) on Flickr

Generative AI is a large language model that learns from large amounts of data, using sophisticated algorithms to recognize and generate new texts, images, video, and audio (multi-modal). At its best, GEN AI should amplify, not replace, human abilities, and it should empower people and ensure human control. Ideally, GEN AI should provide support services to humans with a human-centered focus. Using AI should complement and enhance human abilities, not supplant them, for a collaborative relationship between human and machine intelligence.

Though largely invisible, artificial intelligence has been used in our everyday lives for years. We are continually having common encounters with AI. For example, smart phones--from facial recognition software for unlocking the phone and other apps, to speech recognition to do things like translate from one language to another. Anytime we use a virtual assistant, like Alexa or Siri, it is undergirded by artificial intelligence. Customer service chatbots, wearable fitness trackers, custom product recommendations, and customized music playlists are all supported by AI.

Generative AI represents another step forward in artificial intelligence:

• It creates new content, such as text, images, code, audio and video.
• Generative AI text generators or chatbots, such as ChatGPT, work by predicting the next word in a sequence of words. The majority of text has been scraped from the internet and then supplemented by feedback testing done by humans. Image generators, such as DALL-E, work in the same manner.
• Not much is known about how the models that drive these tools were trained. That means that, in most cases, there is no way to know what information an AI tool has had access to.

Generative artificial intelligence is a revolutionary technology, but it should not be embraced wholeheartedly. It has limitations (at least at this time):

• Some AI generators, such as Microsoft's CoPilot, have access to the internet, but only when using the Bing search index. Free versions of ChatGPT do not have access to the internet, so a query will not result in the most recent information on a topic. ChatGPT Plus is a subscription plan which provides access to the internet and more accurate and reliable results from a query.
• Because most text generators capture and reuse information obtained by a query, any information shared with these programs has the potential to be shared publicly. ChatGPT Edu does not reuse information from queries or or other data provided to ChatGPT.
• All text generators produce seemingly plausible information, but often fabricate information and introduce inaccuracies. When a chatbot makes things up, this is known as an "hallucination." Using a chatbot is not the same thing as using a database or a search engine (like Google). When using these tools, manage expectations. For an explanation of how and why these large language models hallucinate, see Will Douglas Heaven. (2024, June 18). Why does AI hallucinate? MIT Technology Review.
• Generative AI systems often produce culturally and politically biased information.
• Potential to diminish critical thinking, creative problem-solving, and other essential skills.

In late May 2024, OpenAI introduced ChatGPT Edu, a subscription version of ChatGPT built especially for colleges and universities. ChatGPT Edu can review resumes, help with grant applications, and assist faculty with feedback and grading. The Edu version is also a safer version, as no conversations or data is used to train OpenAI models. Although this version is not available at Hood College, it shows how generative AI can be tailored to meet the needs of particular types of organizations or enterprises.