The Chatbot Generation
One of the methods educators have used to engage students about AI and change the narrative around tools like ChatGPT as little more than a task completion machine is to give students opportunities to explore this technology under a guided-use framework. Here’s a tool, here’s what it can do for you, here’s what it cannot do, and here’s the danger of abusing it. While we've been implementing various approaches like this across disciplines, there remains a fundamental challenge—ChatGPT's user experience is designed to provide quick, easy answers—which often undermines genuine learning.
I’ve argued that we need generative interfaces that use AI to support meaningful human interaction rather than replace it. There’s very little interaction a user gets out of a pile of context from a single prompt. The generative model views that material only in a split second during inference, then spits out an answer. Of course, users can iterate with a model—prompt, review output, adjust prompt—rinse, and repeat until they like the outcome. However, rate limits and a user experience that prides speed and efficiency over engagement with messy things like ideas, really limit that experience.
Students Deserve Better Than AI Shortcuts
Google's newly launched Career Dreamer tool charts a much different path for users. Instead of simply generating resumes or job letters, Career Dreamer guides students through an exploratory process, helping them understand how different skill combinations affect career opportunities. It also takes the user through the technical jargon associated with job postings and lets them see how one career might interact with many others. That’s what impresses me most about a tool like Career Dreamer—its nuanced way of helping users by letting AI assist them and not simply generate a resume or job letter, but rather use AI as a guide to parse information.
Career Dreamer could never replace a thoughtful or engaged mentor guiding someone young along a career path, but for students who want to explore what careers their degrees or emerging interests might take them, a tool like Career Dreamer is far more effective than asking someone without any experience to hop on a site like LinkedIn and try to figure out their future.
We’re going to need tools like Career Dreamer along with human beings guiding students in using such tools as starting points, but not automating the complicated parts in the process. Finding a career is a human process, one with a frustrating level of trial and error, productive struggle, and yes, hopefully, some dreaming. If people used an AI tool like Career Dreamer as a problem solver instead of as an explorer, then it would ultimately fail in its intended purpose. That’s not a bad thing. AI developers should consider taking a kitchensink approach to AI in a quest to build something that can reason is really what people want or need from this emerging technology.
While tools like Career Dreamer demonstrate how AI can support meaningful exploration and learning, the tech industry's broader focus often veers in a different direction. Case in point: the latest sensation in generative AI.
The Great AGI Promise Delivers... Animated Memes
The internet is abuzz with OpenAI’s new image generation capability. Posts across social media are filled with users taking images from films and using AI to alter them. The most popular is the Studio Ghibli style of animation and the results are impressive, but ultimately soulless. Two and half years since the public release of ChatGPT we’ve gone from feverish talk about true Artificial General Intelligence just around the corner, to using AI to turn our favorite memes into anime characters. Who wouldn’t feel let down? We once again have an update from OpenAI that causes an uproar and untold numbers of people rushing to use it, with little pause in asking about copyright or the implications for generating deepfakes or energy consumption.
While the text generation found in the version of OpenAI’s GPT-4o is much more efficient than GPT-3 or ChatGPT’s original model, there’s no telling how much energy is being thrown away on memes. How sadly fitting it is that the GPUs powering AI in data centers originally designed for playing games are now being put into overdrive turning social media into a game of who can Ghiblify their network the fastest. AI energy usage will become more efficient over time, but far more challenging problems will persist.
This contrast between generative AI's potential for guided learning and its actual deployment as often an entertainment device or task completion gizmo reflects a broader disconnect in how this technology is introduced to students. The same generation experimenting with AI-generated memes encounters these tools in real-time in educational contexts with far-reaching implications.
AI is a Massive Public Challenge
The crux of the issue with generative AI being released as a public experiment is that developers have chosen to rely on human beings as test subjects, pathfinders, and critics to establish practical use cases and raise alarms about the harms their very tools have created. With ChatGPT crossing over 400 million weekly users, it’s hard to call this choice a failure from a business standpoint, but at what cost?
The majority of ChatGPT’s 400 million worldwide weekly users are students, who have come to use and view this technology as a cheating tool, crutch, or way to save time on school work by offloading learning. The chief worry AI developers should have is how their users view this technology. An entire generation of students will come of age within this generative era. Does anyone believe their exposure to the technology as a cheating tool, deepfake generator, or a creepy simulator of human voices and avatars will lead them to view it positively?
The rhetorical landscape is taking shape for the future of human digital interaction with AI and how real flesh and blood people feel about this technology and consider what it does to their lives now will matter far more than any hyped tool release, think piece, or marketing campaign ever will. Students grow up into adults and decades from now those adults likely won’t look back too fondly on how they witnessed others or themselves use tools like ChatGPT to cut corners in school, harass or be the victim of harassment via deepfakes, or lose their sense of what’s human through conversational AI. Such scenarios have long-term consequences for our world and end up shaping public policy decisions in ways that are often reactive. That’s something we should try to avoid.
Generative tools are shaping our experiences online and offline. It shouldn’t surprise us if behaviors also change. Students who heavily use ChatGPT already exhibit changes in their learning habits, just like heavy users of social media exhibit changes in their attention spans. That can translate to being less likely to challenge ideas, critically evaluate arguments, or deeply engage with topics.
We want our students to understand their relationship with technology is ethically complicated and AI now joins the pantheon of digital things we worry young people will struggle to navigate without assistance. Without practical guidance about how best to use or avoid using AI, students will be left on their own to find their way through an increasingly dark forest of AI tools they hear about from their friends and on social media. That’s what happens when you unleash a technology on the world without a user guide or best practice framework for more things other than “here’s how you can prompt.”
I teach high-school English to good kids. But the temptation is just too great for them—and they represent the public at large for whom bread and circuses suffice to guide daily choices. No amount of ethical preaching will work. Regulation (many different guises for this, like local LLM’s and training in how to use them) or a commitment to offline learning is the only way forward.
The potential societal impacts are terrifying. We already have filter bubbles, now we'll have bespoke personas - so much easier to get along with than real people. Forget fake news - fake your own news, and have it delivered by your favorite flavor of cartoon character! It is horrible to be living in the genesis of the Mad Max/Terminator/zombie apocalypse but powerless to stop it, knowing I won't live long enough to see the beginning of the movie, much less the part where the hero/es start to turn the tide.