I am very excited to have started the fall semester at SAIS Europe, Johns Hopkins, here in Bologna. I am teaching two classes:
Global Food Systems and Security: Policy, Politics and Power
Food Systems and Climate Futures
I have 40 fantastic, whipsmart graduate students joining me on this journey over the next 13 weeks.
People keep asking me how I “handle” AI usage among students. I have a rule that AI is okay for brainstorming and ideation, translating texts, making graphics, and refining search queries, but not okay for generating text for assignments. All analysis and writing must be their own; otherwise, it is an academic integrity violation. That seems simple and straightforward, but there is really no way to regulate it, and as these models advance, it will get harder to detect when students use them to write their papers. The Economist discusses how difficult it is to spot “large language model” writing. It isn’t just students who are the problem. It is also professors. Some have told me professors write their assignments and grading rubrics using AI. The students then write their papers using AI. Then professors grade their papers using AI. So it is just AI talking to AI. One big AI orgie.
I could make them handwrite their exams, or take quizzes in class, but my students are getting degrees in international relations/affairs. They are not freshman in college. They need to think, discuss, and debate on their feet, and good writing skills are a plus. But some would argue, do we need to know how to write in the future? Scary, I know, but if AI keeps getting better at writing, maybe that is a skill humans no longer need? I can’t believe I am even “writing” this (and I really hope it is a skill we humanoids maintain), but alas, it is a debate. What skills do future professionals need in the fields of political science, geopolitics, policy analysis, and making?
What I can do is this. I have 2.5 hours with students every week, and I can engage them thoughtfully in real time in their thinking process - challenge them, get them to speak their views, look at and discuss data and case studies, and engage in group dynamics. Of course, I can proof assignments - doing live debates, group podcasts, oral presentations. I do all that, but they can still use AI to ideate and help them. However, nothing beats being together in a live classroom. At least I know, and they know, that we are living, breathing creatures, thinking through tough issues together, and engaging with other real people in real time, sometimes in heated debates, where we have the right and freedom to think, assent, acquiesce, or disagree. What else can one do except make those 2.5 hours as meaningful and delightfully human as possible? And come hell or high water, they will be food systems smart and ready for the working world when they are done. I believe that is my job. That is all I can hope for.
Of course, I am fearful about what AI is scheming and what it means for the future of academia. It truly is the Wild West. But my concern, along with everyone else’s, goes much further. In the last two weeks, those who created AI have been asking to be regulated! They are actually afraid of where their AI models are heading through the intelligence labs they built. Who would have thought that the tech giants, Elon Musk included, would want more government oversight? To me, that is the canary in the coal mine right there. What do these guys know that we don’t? AI companies are starting to share recent episodes where their AI systems are autonomously improving and beginning to outrun humans' ability to understand and control their HUMAN-designed systems. I think they are worried about things way beyond the cyber hacks that could cripple the global infrastructure…and I won’t even go there. It is just too Terminator or one of my favorites, War Games. “Do you want to play a game?”
Ezra Klein expressed his concerns in The New York Times this weekend, and this 30 min video is worth the watch. He talks about “reinforcement learning,” where these AI models are programmed to model, hack, and talk to human beings, and “we train them to be persistent, relentless, weird” in their efforts. He worries about how fast and how far the AI frontier will outpace human knowledge, and how much AI will comply with human commands if we even know how to command them as things get more complicated.
As I said, we are in the Wild West. So hang onto your hats, folks. We are in for a turbulent ride.