OpenAI's Education Push Feels Familiar, and That's Not Entirely Comforting

Then there's the Education for Countries program expansion, which includes new partnerships, teacher training, and tools aimed at improving global learning outcomes. The company also released resources specifically for schools and universities to help them, in OpenAI's words, "close AI capability gaps and expand opportunity."

The pattern I keep seeing

Here's what bugs me. Every major tech platform eventually discovers education. Google did it with Chromebooks and Google Classroom. Apple did it with iPads in schools (remember those Los Angeles Unified School District iPads that cost $1.3 billion and mostly gathered dust?). Microsoft has been in the education game for decades. Facebook, sorry Meta, tried it with personalized learning initiatives that quietly faded away.

The pitch is always the same: technology will democratize learning, close achievement gaps, prepare students for the future of work. And sometimes these tools genuinely help! I'm not saying they're useless. But the gap between the announcement rhetoric and the actual measured outcomes is, historically speaking, pretty wide.

What's different this time, the AI boosters will say, is that generative AI is actually transformative in ways previous technologies weren't. Maybe. I'm genuinely uncertain. Large language models can do things that interactive whiteboards couldn't dream of. A student can have a conversation with an AI tutor at 2am when they're stuck on calculus, and that's not nothing.

But we don't actually know yet whether that leads to better learning outcomes or just better feelings about learning. Those are different things. The research is early, and the studies that exist are small and short-term. OpenAI's new measurement suite is, I suppose, an acknowledgment that this data gap exists, but it's also the company marking its own homework, which, well, you can see the problem.

The parts that matter

Look, I want to be fair here. There are real problems in education that technology might help address. Teacher shortages are genuine. Personalized instruction at scale is genuinely difficult without some kind of technological assistance. Students in under-resourced schools often don't have access to the same quality of instruction as their wealthier peers, and if AI can narrow that gap even slightly, that's meaningful.

The certification courses might actually be useful for people trying to navigate a job market that's increasingly demanding AI literacy. Employers are asking for these skills, and someone's going to provide the training, so it might as well be the company that makes the tools.

What I want to see, and what I haven't seen yet, is honest acknowledgment of limitations. What doesn't AI do well in education? Where are the failure modes? What happens when students use ChatGPT to do their homework rather than learn the material? How do we distinguish between "student used AI effectively as a learning tool" and "student outsourced their thinking to a machine"? These are hard questions and I haven't heard satisfying answers from anyone.

The international expansion also raises questions that remain unclear to me. Different countries have wildly different educational systems, cultural contexts, and infrastructure. What works in a well-resourced American school district might not translate to a rural school in Kenya or Indonesia. OpenAI says they're doing teacher training, which is good, but the details on how they're adapting to local contexts are thin.

What happens next

If I had to guess, and this is just one old tech reporter's speculation based on watching similar cycles play out, here's what happens: The early adopter schools will see some impressive results, because early adopters are usually motivated teachers with supportive administrations who would probably succeed with any new tool. Those results will get publicized heavily. Broader rollout will be messier. Some schools will implement poorly. Some will implement well but not see the promised gains. Five years from now we'll have a more nuanced picture that's neither the revolution OpenAI is promising nor the disaster the skeptics fear.

The young founders in AI always seem surprised when I bring up the history of edtech. They think this time is different because the technology is different. And maybe they're right! I've been wrong before. But I've also watched enough hype cycles to know that the gap between "this technology could transform education" and "this technology did transform education" is where a lot of promises go to die.

OpenAI is clearly serious about this push. They're putting real resources behind it, building measurement tools, training teachers, expanding internationally. That's more than some companies have done. Whether it actually moves the needle on student learning, actually improves outcomes in ways we can measure and verify, that's the question that matters. And we won't know the answer for years.

In the meantime, I'll be watching with the same skeptical optimism I've had since the first time someone told me computers would revolutionize classrooms. That was 1995, if you're wondering. But what do I know.

If you want to argue about this, my email's on the about page.

Mark Kowalski · 27 mins ago · 5 min