A generation is growing up with AI as the final word.

Professionals who once relied on their training and their colleagues turn to a model first. It speeds up how fast we learn, but it builds a dependency, much as reading once did.

The same conflict Plato saw is brewing again. Reading accelerated mankind, enabling the Scientific Revolution and four centuries of progress. But it changed us. Our relationship with information became text-based. Our memories are translated through words now, and, as Plato feared, we use our own memory less. Has the human mind atrophied in some sense? Perhaps. Once something is written down, remembering it is no longer necessary. A new version of this is happening with AI. As we lean on the model instead of our own judgment, we lose the habit of checking it. We take what it gives us. Our relationship with knowledge and fact is being redefined again, this time through what the models feed us, and we are less and less in a position to notice when they are wrong.

This could move science and industry forward, but only if the models tell the truth.

Online data only goes so far. Passing accurate knowledge to the next generation takes real human expertise, captured now, before the takeoff. So we sit with the people who hold it. We watch how a specialist actually reasons, the calls they make and the ones they reject, and we turn that into data a model can learn from. Doctors first, then the fields after them.

Leave it too late and the slop gets passed down instead: junk medical knowledge fed to future doctors, fake practices described to new law students, wrong maintenance instructions handed to refinery workers. These errors permeate, and the generations after us will scramble to detach hallucination from fact.

Train models on documented best practice and you give the minds of the future something solid to build on. That is where we start.