Since AI seems to help the least-educated most, unequal access to it costs more than it first looks.

This explores whether people with less formal education get the same chance to learn and benefit from AI tools as people with degrees. The honest answer is that the collection doesn't measure who actually receives AI training by education level. What it does show is more surprising: the people with the least formal education seem to gain the most from AI. That makes unequal access more costly than it first looks.

The strongest evidence comes from a randomized experiment with more than a thousand adults. Generative AI closed about three-quarters of the performance gap between more- and less-educated participants on a business problem-solving task. Less-educated participants also kept part of their improvement after the AI was taken away Can AI narrow the education performance gap?. That last detail matters. It suggests AI use worked partly as training, not only as a crutch. Hiring shows the same pattern. In one experiment, listing AI skills on a résumé partly or fully offset the interview penalty for candidates with associate degrees instead of bachelor's degrees. The effect was strongest for office assistant roles Can AI skills help older or less-educated job candidates?. So AI skills may be one of the few credentials that can stand in for a degree.

The broader labor market pulls the other way. Job-posting data from ten countries shows AI skill demand gathering around a technical core of Python, SQL, machine learning, and data analysis inside STEM jobs. Non-technical occupations are moving away from that core, not toward it Is AI creating common skills across jobs or deepening divisions?. If 'AI training' comes to mean that technical toolkit, it will mostly reward people who already have technical education. Who is exposed to AI is also uneven. In female-dominated occupations, exposure reaches all skill and wage levels, so lower-paid women face disruption with fewer resources to retrain Does AI exposure hit low-wage workers harder in some fields?.

The common thread is that the technology doesn't decide the outcome. An interdisciplinary review across work, education, and healthcare found that generative AI can either widen or narrow inequality. Which way it goes depends on who gets access, how the tools fit into daily work, and what incentives are in place Does generative AI inevitably worsen or reduce inequality?. A more philosophical note adds another layer. These models are built from humanity's shared written output, so restricting access to them turns a collectively produced resource into a private advantage Should restricting AI access create new kinds of inequality?.

One related idea comes from model training itself. When a large 'teacher' model creates training material for a smaller 'student' model, the student can get worse if the material is beyond what it can currently learn Does teacher-refined data always improve student model performance?. That research is about models, not people, so treat it as an analogy rather than evidence. Still, it hints that equitable AI training may not mean giving everyone the same course. It may mean meeting learners where they are. The takeaway: the evidence suggests less-educated workers have the most to gain, and the open question is whether training will be designed around them or around the technical core.