What to Read about Generative AI

Stay up to date on generative AI discourse!

Posted on July 31, 2024 in Blog Posts

Authors:

Daniel Pfeiffer
A librarian reading about generative AI and ChatGPT

Generative AI news and reports form a perpetual froth on the surface of the internet. Given its ceaseless multiplication, you would be forgiven for tuning it out. The past few months haven’t produced many major stories. Lawsuits against major tech companies continue to work their way through the system but haven’t made any headline-grabbing stories. But outside of the spotlight, some interesting articles have emerged, and this week, we have you covered with five pieces we think you should read. 

🕵️ “Can AI detectors save us from ChatGPT? I tried 6 online tools to find out,” David Gewirtz (ZD Net)

AI detection software is a polarizing issue. On the one hand, educators want a means of enforcing academic integrity, knowing that AI tools equip students to cheat easily. On the other hand, some worry that detection software creates an atmosphere of paranoia, straining the bond between educators and students, especially given the potentially severe consequences of false positives. What defense do students have against false accusations of cheating?

Regardless of where you stand on this issue ethically, it’s important to have a technical understanding of this software, its accuracy, and its limits. Gewirtz tested six services using a combination of human- and AI-generated text. Though three services were correct 100 percent of the time, he concludes that he “would not be comfortable relying solely on these tools to validate a student’s content.” His piece includes a lot of details of his test and screenshots of the software interface, making it a good primer on the current selection of AI detection tools. 

🎓 “AI Has Hurt Academic Integrity in College Courses but Can Also Enhance Learning, Say Instructors, Students” (Wiley)

I’m normally skeptical of “feelings” surveys about AI. Whether 30 percent or 70 percent of students feel excited about AI doesn’t change the fact that we need to think about how we should educate students about AI. However, I make an exception for this new survey out of Wiley because its sample size is larger than most, encompassing 850 college instructors and 2,067 college students in the US and Canada. Its findings are predictably mixed: both groups see some benefit to AI and some harm.  

Its most interesting point is that cheating hangs heavy in the discussions and uses of AI. Many students don’t want to use AI at all because they’re afraid that it would constitute cheating, and instructors are similarly preoccupied by how to limit cheating. Though I recognize the fear of creating an atmosphere of paranoia by policing students’ AI uses, I think this survey tells us that we can’t sidestep the cheating issue. Universities and instructors will have to develop clear guidelines to set these boundaries. 

🌟 Subscribe to the LibTech Insights newsletter for weekly roundups and bonus content, including: 

💸 “AI & Universities,” Mary Meeker (BOND)

This report requires a bit of context. Its author, Mary Meeker, is likely unfamiliar to academics, but Meeker is a big-name venture capitalist and the founder of BOND, the venture capital firm that published her report. Her Internet Trends Report is always hotly anticipated by tech investors. This year, she decided to focus on universities and AI. To editorialize a bit, her report is filled with a lot of business jargon (and pablum—“the needs of a rapidly evolving marketplace supercharged by innovation”) and rehearses many familiar points about the potential for AI to affect education (individual lesson plans, teacher support, job training). Less important is what Meeker writes than the fact that her eye and therefore, venture capitalists’ eyes are on universities as a primary arena for AI tech investment—for better or worse.

For more analysis and reactions, check out this article in Inside Higher Ed. 

🌳 “Generative AI requires massive amounts of power and water, and the aging U.S. grid can’t handle the load,” Katie Tarasov (CNBC)

Often lost in discussions about the ethics of generative AI is its profound energy costs. Servers require large amounts of energy to run and be cooled, and generative AI requires large amounts of servers. This article outlines the energy problems AI is incurring and the problems of the “solutions.” It includes some startling statistics: Google’s greenhouse gas emissions rose nearly 50 percent from 2019 to 2023 and Microsoft nearly 30 percent from 2020 to 2024, in part due to data centers. Not only are the energy costs dangerous to the environment, but they are also taxing the already precarious US energy grid and may soon push it closer to its breaking point. 

📚 “Academic authors ‘shocked’ after Taylor & Francis sells access to their research to Microsoft AI,” Matilda Battersby (The Bookseller)

In scholarly publishing news, Informa, the parent company of Taylor & Francis, signed a deal with Microsoft, granting it access to its data to train its AI models. The problem? It didn’t alert authors of this deal … or give them the possibility to opt out … or cut them a slice of the £8 million ($10 million) agreement. The exact details of how these data will be used to train Microsoft’s AI models are unknown, and Microsoft was cagey when asked for comment. However, authors should heed James McConnachie’s call in the article to ask publishers about their AI policies and look carefully at their contracts.

🔥 Sign up for LibTech Insights (LTI) new post notifications and updates.

✍️ Interested in contributing to LTI? Send an email to Daniel P. at Choice with your topic idea.

Related Posts