Jonah Goldberg Argues Against AI in American Classrooms

Columnist Jonah Goldberg published an opinion piece in the Las Vegas Review-Journal arguing that artificial intelligence should remain largely absent from American classrooms. The April 25, 2026 column draws a direct parallel between Norway's 2016 decision to distribute iPads to every child starting at age 5 and the potential consequences of introducing AI chatbots into education systems.

Goldberg's central claim rests on a specific data point from the Review-Journal article: approximately 500,000 Norwegians cannot read a text message or simple instructions. The column cites The Times of London for this statistic and references the Progress in International Reading Literacy Study (PIRLS), which reportedly ranks Norway at the bottom of 65 countries for children's enjoyment of reading.

These numbers warrant scrutiny. A population of 5.6 million with 500,000 functionally illiterate adults represents roughly 9 percent of the total population. That's not a rounding error. Norwegian Prime Minister Jonas Gahr Store reportedly launched a program in August to address the problem, stating that 15,000 pupils finish primary school without being able to read properly. The timeline here matters: the iPad distribution began in 2016, and the literacy crisis emerged a decade later. (That's a long lag time for policy consequences, but not unprecedented in education reform.)

Goldberg's argument extends beyond the Norwegian case study. He frames education as "basic training for civilization," comparing it to military preparation where personnel must master fundamentals before operating advanced technology. The physical reality of this comparison is worth noting: a soldier doesn't learn to pilot a drone by reading about drones. They learn through repetition, failure, and gradual skill acquisition. Goldberg contends AI removes this essential friction from learning.

The column makes a specific distinction between adult and child AI use. Goldberg acknowledges merit in AI for "existing highly skilled workers" who can leverage it to increase productivity. But he questions how those workers became skilled in the first place: by doing the work. This creates a logical tension in the broader AI-in-education debate. If AI is a tool for augmentation, when does augmentation become substitution? The answer depends entirely on what educators define as the learning objective.

Goldberg draws another historical parallel to calculator adoption in the 1970s. Educators learned that students needed to master mathematical fundamentals before delegating computation to machines. He describes AI as "essentially a souped-up calculator for nearly all mental tasks." The comparison has limits, though. Calculators perform arithmetic operations with deterministic outputs. Large language models generate probabilistic text that can be persuasive, creative, and occasionally wrong. The cognitive engagement required to verify AI output differs substantially from checking a calculator's math.

The physical experience of learning with AI versus without it reveals the core of Goldberg's concern. When a student types a prompt into a chatbot and receives an answer within seconds, the friction of research, synthesis, and composition disappears. There's no dog-eared textbook, no scribbled notes in margins, no frustration of searching for the right source. The learning happens in the struggle, not the result. Goldberg argues that removing the struggle removes the learning.

Proponents of AI in education use terms like "cognitive shift" and "upskilling" to describe this transition. They argue that by removing drudge work, students can focus on higher-order thinking. Goldberg counters that you can't reach higher-order thinking without first building the foundational skills. It's a sequencing problem. The question isn't whether AI has value in education, but when that value becomes accessible without undermining prerequisite competencies.

The column's policy prescription is blunt: education should "mostly stay in the pre-AI world." Goldberg acknowledges this will be difficult. It requires more memorization, more in-classroom testing, and an education establishment capable of resisting "technological fads." The last point carries particular weight given the history of ed-tech adoption. Schools have repeatedly embraced new technologies with enthusiasm, often without sufficient evidence of learning outcomes. The iPad rollout in Norway serves as a cautionary tale, but it's not the only one.

Goldberg's conclusion offers a nuanced position rather than outright rejection. He argues the point of education in the AI era should be equipping students to ask the right questions, including the ability to question why an AI chatbot gave specific answers. This shifts the learning objective from answer-finding to answer-evaluation. It's a different skill set entirely, one that requires students to have enough foundational knowledge to recognize when AI output is reasonable or flawed.

The broader context matters here. This opinion piece arrives during a period of rapid AI integration in schools. Some districts have already begun implementing AI tutors, automated grading systems, and personalized learning platforms. The debate isn't theoretical anymore. Schools are making decisions now about whether to embrace or restrict these tools. Goldberg's argument adds to the growing chorus of voices questioning whether efficiency gains justify potential long-term costs to cognitive development.

Whether schools actually heed this warning remains an open question. The technology is already here, and the pressure to adopt it is substantial. Parents, administrators, and policymakers will have to decide what they're willing to trade for convenience. The Norwegian example suggests the consequences of getting this wrong could take a decade to fully manifest. By then, the damage may already be done.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn