Let’s Look At This Clearly: AI Hype in Ed Tech

A note from September 2026: I wrote this in fall 2024 and have left the argument as it stood then. The critique of hype still holds, but my view of where AI belongs in education has narrowed since. For where that went, see Second Thoughts on AI-Assisted Game Design in K12 and the Vibe Coding in EdTech series.

Over the past year I’ve had a front-row seat to the hype around artificial intelligence. Tools like ChatGPT and DALL-E can be useful in specific circumstances, but the breathless speculation and grand claims about what they can do are creating more confusion than clarity, and schools are where a lot of that confusion lands.

Where is the hype coming from?

A fair amount of it comes from the companies themselves. Ed Zitron, in a piece asking whether we’ve reached peak AI, points out that OpenAI’s Sam Altman tends to drift into what Zitron calls “fan fiction,” talking about everything AI might someday do without naming a real use for ChatGPT today. Product launches follow the same pattern. In a Wall Street Journal interview about Sora, OpenAI’s video generator, then-CTO Mira Murati couldn’t give basic details about the tool and kept falling back on capabilities that would arrive eventually.

The media doesn’t help. Zitron argues that reporters have been taken in by AI much the way they were by the metaverse, and that the half-truths spread faster this time because AI actually exists. It’s easy to picture it changing our lives, even when the specific changes being promised range from unlikely to impossible.

What does that look like in schools?

In my work supporting districts with technology integration, I’ve watched this hype cycle turn into misunderstandings about what AI tools can do. Educators get excited about AI transforming their classrooms and are then let down when the reality turns out to be much narrower. The misunderstandings I run into most are 1) overestimating how well AI can personalize learning, along with how much that personalization is worth; 2) expecting it to take over grading and feedback; and 3) confusion about its role in content creation and plagiarism detection.

Part of the gap is that current models, however fluent they sound, don’t reason or understand in any meaningful sense. In a recent paper, Michael Townsen Hicks, James Humphries, and Joe Slater argue that large language models do far less than human brains do, and that their main aim, to the extent they have one, is producing text that reads as human. They call that output “b*******” in the philosophical sense, meaning text produced without concern for whether it’s true. That sounds harsh, but it gets at how differently these models work from human thinking.

Is there a balanced approach?

There is, as long as the tools are used for what they can actually do. When I present to educators around the Midwest, the uses I point to are modest: 1) early content curation, resource development, and formatting; 2) generating extra ideas for discussion prompts and activities; 3) basic writing assistance; 4) answering general questions about a document or knowledge base; and 5) acting as a simple coach or roleplay partner. Each of those comes with limits and pitfalls, and I spend as much time on those as on the uses themselves.

The hype makes that harder, in that it wears down AI literacy for educators and students alike. When people keep hearing that superhuman AI is just around the corner, they lose track of what current tools actually do, and grounded conversations about appropriate use get harder to have. AI literacy in schools needs to cover how to evaluate AI-generated content critically, the basics of how the models work, the ethical questions they raise for education, and what responsible use looks like in practice.

Where does science fiction come in?

At ISTE Live 2024 I presented on how science fiction shapes our expectations of generative AI, looking at three familiar types: the evil AI like HAL in 2001: A Space Odyssey, the flawless assistant like Jarvis in Iron Man, and the human-like AI of Her. Those stories feed straight into the hype cycle, and my point in the talk was that educators need to think about AI critically rather than through pop culture. I wrote up the fuller version of that argument in How Science Fiction Can Improve EdTech & Design.

So where does that leave us?

Zitron describes each new round of claims about what AI can do as somebody adding fuel to a fire that’s close to going out. I’m not sure the fire is going out, but schools don’t have to keep feeding it. What’s left is slower work: figuring out what these tools actually do, where they fall short, and helping educators and students tell the difference.

If you’re sorting through AI claims in your own district, I’d like to hear what you’re running into. You can reach me at licht.education@gmail.com, and there are more tools, articles, and resources at bradylicht.com.