Generative AI

Generative Artificial Intelligence is just one type of artificial intelligence, and it’s important to know what it is and how it was developed. First, unfortunately, there isn’t one official definition of “artificial intelligence.” NASA offers a straightforward one: “Artificial intelligence refers to computer systems that can perform complex tasks normally done by human-reasoning, decision making, creating, etc.” If we think about the human brain as a machine or a network, then it has seemed possible that a machine could be created that could mimic the brain.

Generative AI refers to networks that generate words, images, voices, videos, computer codes, and such, based off of databases full of words, images, voices, videos, computer codes, and such. After these words, images, etc. are collected into a database, a large language model (LLM) is created (to generate text) or a a multimodal foundation model (MfM) is created (to generate texts, images, audio, code, etc.).

The New South Wales government website offers a helpful breakdown of the major types of artificial intelligence: generative AI, machine learning, natural language processing, and computer vision. The University of South Florida Library has a beautiful timeline visualization with some of the biggest moments in the development of AI.

But how does a chatbot like OpenAI’s ChatGPT work? Imagine that you could collect all of the words, phrases, sentences, images, music, and computer code off of the Internet, from all the nooks and crannies (including some of the nastiest parts of the Internet), and dump them them all into a data base. The you constructed a probability model (LLM) that could be trained to predict, based on percentages, what word, phrase, image, etc. should go in what order. Based on the immense amount of data, and a lot of training, the model could get better at predicting what order the data should go in so that it made sense (so that it was, perhaps, a human-sounding answer to a typed question). The model does not “think” or “know” anything; it works off of probability, and it is more likely to put the data in the right order if it has a whole lot of data and a whole lot of training to get the correct weights put on said data.

This explanation by Stephan Wolfram is rich and in depth if you really want to get into the nitty gritty: “What Is ChatGPT Doing … and Why Does It Work?”

Because of the natural language processing function, a chatbot like ChatGPT produces output that sounds like a human was typing back to you. It’s not “typing back to you;” neither is it “thinking.”

The discussion about what AI literacy is and is not is a hotly debated topic. The  SUNY Librarians Association Advancing AI Literacy Working Group has just published SUNYLA SILC: Advancing AI Literacy, which covers AI literacy from an information literacy framework. My approach to generative AI literacy is slightly different: I think people need to first have an understanding of what generative AI is and how it was developed before they start engaging with it. I know that may seem like an unnecessary first step, especially for folks who are not interested in computer science. We don’t interrogate how Elsevier products are created, for example (maybe we should?). All I can say is that, in my experience, when people come to understand a fuller picture of how we got to this point with these generative AI products, that knowledge informs their approach to generative AI.

Therefore, it is ethically and morally sound to make sure students can make informed choices about if, when, and where they might encounter and where they might encounter and where they might encounter or use generative AI, based on a general understanding of what generative AI is and what it is not. To that end, this is my list of student learning outcomes (SLO) for a robust generative AI literacy

​Generative AI Literacy SLOs

• Explain a brief historical context of AI and machine learning ​
• Explain how this current iteration of generative AI was created ​
• Explain the environmental and human impacts generative AI perpetrates ​
• Explain the effects of interacting with generative AI on cognition ​
• Assess the contexts where people interact with generative AI​
• Evaluate use cases for generative AI and limitations in those use cases ​
• Assess how generative AI is being used in eventual (or potential) workplaces​

Module 1: History and Context

Module 2: Generative AI Today

Module 3: Environmental and Cognitive Impacts

Module 4: Case Studies and Research Tips

There are types of artificial intelligence that have been around for many, many years, in many different industries. They are not energy intensive, they are unobtrusive, and they operate under very specific parameters. I am not concerned about them.

When I talk about “AI,” and the relationship between AI and data centers, I am talking about generative artificial intelligence being fueled by gigantic data sets on which large language models train.

That is, generative artificial intelligence that is used to:

• Replace teachers and tutors
• Steal from artists and authors to create slop
• Replace friends and therapists
• Impersonate health influencers to share dangerous health information
• Create Child Sex Abuse material that is almost indistinguishable from real CSAM
• Encourage people to harm themselves and others

And what is the relationship between this kind of AI and data centers? Companies need giant computers to run those large language models, and it takes a tremendous amount of energy and water to keep those machines running. They produce noise and pollution, for the benefit of temporary jobs, many of whom will be filled by out-of-town contractors.

We must continue to lead the country with our resistance to data centers. People are being inspired by us, and we are giving them precious hope and strategies to follow. Our work continues, though — nothing is set, nothing is final, and I think of these developers like insulation foam: they push into communities until they hit a wall, like YOU, protestors, and politicians with spines and a commitment to the health and well-being of their communities. But the foam does not stop — it just diverts and keeps pushing on weak spaces to invade. Now, in the basement of my house, this is awesome! In this analogy, it means our neighbors need us now to continue this work of education and resistance.

This work is essential to me as an educator and a writer. As the director of a community college writing center, I see generative AI as an existential threat to my students. There is nothing more sacred to me than helping students learn to write, to think critically, to reason and explore language and composition, and generative AI undermines all of that. My advocacy on behalf of my students demands that I fight on every front to keep threats to their critical thinking abilities at bay.

Finally, I am seeing a Reply Guy in almost every social media post regarding data centers, who posts something along the lines of, “If you don’t want data centers, then stop all the mobile phone use, video games, Internet, etc. Because you have to have data centers for all that data.” First, don’t threaten me with a good time. We are seeing a return to analog media and a nostalgia for times like the early 90s, as popularized in works like Ryan Murphy’s “Love Story: John F. Kennedy Jr. and Carolyn Bessette.” Second, we don’t need to have hyperscale data centers to fuel mobile phones, video games, and the Internet that haven’t been overloaded with generative AI functions we didn’t ask for. We don’t want generative AI apps that create CSAM deepfakes. We don’t want CoPilot to be forced on us, and we don’t want Google to slap AI slop above the sponsored ads to hide poor search results from us. We can determine the future of entertainment and business if we are willing to keep standing up and fighting back.