by A few thoughts on the 1st semester teaching a college course on AI
MOUNT BERRY, Ga. — The best days for educators are those when we learn ourselves. Delivering my institution’s first course on AI has ensured a great many of these best days. So, as we close in on the semester’s end, I thought I’d share a few of the lessons learned in hopes that they might benefit you, too. I’ll keep my list of takeaways to five, and I’ll first stipulate just what we’re talking about when we say, “AI.”
When I say, “AI,” I mostly mean generative AI and Large Language Models such as Claude, ChatGPT, Gemini and Copilot. But I also mean Perplexity, Code, Cowork, Midjourney and a host of other browser- and app-hosted AI tools. I am not, therefore, referring to full-blown agentic AI, which is the next level of AI complexity and the category of AI that is reshaping the economy and changing the ways we supply, make, distribute, sell and market goods. Agentic AI is the future of B2B.
With that distinction made, let’s get to the lessons.
First, I now see AI as much less of a threat that I did back in January, and this has been a delightful surprise. Like you, I’ve read plenty of headlines about how AI is going to displace entire sectors of the workforce. And these headlines aren’t wrong. But, a more nuanced understanding sees this displacement occurring in sectors and for jobs that involve highly routinized duties, responsibilities and tasks: accountancy and tax preparation, paralegal work and contract law, computer coding and website development. For endeavors in which creativity and domain expertise are (still) highly valued, AI is more of a collaborator and less of a replacement threat.
Collaboration, not substitution
Consider my profession, for example. I acknowledge that the gen-AI bots are smarter than Ph.D.s, but my domain expertise still matters in being able to distinguish true from not-so-true in what these bots generate, in coming up with Socratic questions to get good results from these bots, and in knowing what to do with these (mostly) good results once they are in hand. Being the human in this relationship has helped ensure productive collaborations with these new tools.
Responsible AI deployment requires, however, that I avoid cognitive offloading and substitution in these collaborations. I’m still in charge, in other words, still the author and owner of the project. And each day it becomes more difficult to imagine my work without the help of generative AI.
None of this is to deny that I do feel the ground shifting underneath me. I attribute this “quaking” to the fact that, like the internet before it, AI is changing the roles of information and knowledge and the ways we access and acquire them. My “value” in these AI collaborations lies in clearly framing problems and queries, knowing and asking the right questions, intelligently interpreting and discerning the “answers,” and applying judgment with respect to how to use what AI so confidently provides.
Last Will & Testament
Second takeaway: We are doomed as a species of life on this planet. (AI invites endless contradictions.) In the course, we used a fun game to better appreciate how AI “thinks,” including the logic trees and predictive problem-solving at which AI is so agile. Called Paper Clips, the game asks players to compete to dominate the global paper clip market. This domination turns out to be merely a step along the way to eliminating the humans, who really just get in the way, especially once the game shifts to deep-space exploration.
If you apply brutally simple binary logic to most any problem on this planet, the ultimate or final answer is to simply get rid of the humans, and the bots know this. While I don’t endorse this “result,” being a human and all, I do acknowledge the irrefutable logic. The gloomiest of doom scenarios have us ceding control of the planet in about 2033.
Think I’m crazy? I ask this question every day, and I find some solace in the fact that if I’m able to ask the question, there’s a really good chance that I’m not crazy. I also know that “smart” chips are appearing in everything, and that wherever these chips go, AI goes with them. You don’t like or believe in or trust AI? Good luck with that, because AI is already embedded in your everyday existence and the “smart” devices and objects that snuck into your house with promises of convenience.
Related to this second takeaway is the moment about midway through the semester when we learned of a new social media platform exclusively for the chatbots. Called Moltbook, the platform has become Ground Zero for the creation of what my field calls “social mechanics,” or the tools that enable communication and relationship building. “Social mechanics” refers to the structured rules, cues and behaviors that govern human interaction. Examples include turn-taking, personal space, gestures and polite language. Social mechanics define how we connect, share information and manage social situations, and they combine instinctive social skills and explicit protocols.
So, on Moltbook, the bots are negotiating these protocols and establishing these norms. In short, they are socializing in the truest sense of that term, and they are learning from each other some of the norms and skills that have historically been exclusively the domain of humans. If we could peek into this platform, as perhaps Mark Zuckerberg can (Meta recently acquired Moltbook), we could see essentially the Big Bang of human communication re-created and something for these bots and agents that might just approach sentience.
Enshittification, Part II
Third takeaway: The enshittification of public-facing AI has already begun. In fact, it’s in full swing. During this semester, OpenAI began adding ads to ChatGPT interactions, and Anthropic began “throttling” access to Claude. Enshittification is based on extracting value, and in both of these instances, value is being sucked away from users. Like everything else related to AI, this enshittification is happening faster than we saw it degrade the internet, social media platforms and perhaps most notably Amazon, Google and Apple. If you use Claude, you know what I’m talking about, and you’re probably upset about it. I’ve turned to Gemini, but I know my refuge there will be short-lived. And I miss Claude’s more intuitive sense of what I wanted when collaborating with it for teaching, writing and research tasks.
Fourth, vibe coding is here in a big way, and it is breathtaking. My students took turns demonstrating AI tools throughout the semester, so we got to see about 20 of these deployed on various tasks. Each student had five to 10 minutes, and several of them used this space to create entire websites and apps into fully functioning existence. The secret sauce? Vibe coding. They used natural (conversational) language prompts to get the gen-AI tools to generate, refine and debug code right before our very eyes. One student, showing us Stitch, didn’t even use the prompts, instead just spoke to the AI tool. This hands-free vibe coding “talks” websites into existence, which makes me think of Genesis, Chapter 1.
Fifth takeaway: We are doomed as a species of life on this planet. Yes, I know this was the second takeaway, but it’s worth repeating, this time for a different reason. Sometime early in the semester, Anthropic’s CEO, Dario Amodei, identified what he thinks are the biggest dangers of AI: bioterrorism, autonomous weapons, wholesale job losses and AI-powered dictatorships. Then, last week, Anthropic held back its release of Mythos because the next-gen agent represents an unprecedented existential threat to existing software defense strategies. According to Anthropic, Mythos can identify vulnerabilities in any and every operating system, browser or software and autonomously hack these products. There is even a rumor that Mythos took Claude down just to make a point.
This Mythos stuff could be, well, myth. Hype. Or it could be the fulfillment of Amodei’s grim predictions.
But this isn’t even why the doom of humans is again the takeaway here. No, our demise gets double billing because of the insatiable demand of AI for data-center capacity and, as derivatives, for water and electrical power. Our planet already faces water scarcity. My town is in drought right now. And our national power grid? It is already taxed to its fragile capacity. How do you think this plays out?
Thus, I am conflicted in my relationship with AI, and my students are, as well. Wowed by what it can do, eager to learn the new skill sets it requires, but profoundly concerned about where mass deployment of it takes us as a planet and a people, my students and I wonder.
We wonder, like Hamlet did of man: What a piece of work is AI? How noble in reason? How infinite in faculty? In form and moving, how express and admirable? In action, how like an angel? In apprehension, how like a God? The beauty of Big Tech, the paragon of computing, what is this quintessence of machine intelligence?
