Archaeology
Archaeology has a problem in that the data in the field is scattered across site reports, journals, museum records, excavation notebooks, radiocarbon logs, GIS layers, old drawers, and PDF scans. Each fragment is trapped in its own idiosyncratic format.
Archaeologists are famous for over-interpreting the little data they have access to. Dreamers mostly. The interpretations proliferate because no one cares enough to stop them and there’s no consequences for being wrong.
Imagine instead that archaeology decided to do something radical and tedious: build a fixed, standardised database format and force every new and old record into it.
Stratigraphy, dating results, materials, provenance, confidence levels, references. A schema that is strict enough to make data cumulative. It would take years of work, but it would create a workable database of knowledge instead of graveyards of disconnected stuff.
What would be excluded is the last few hundred years of speculative interpretations.
At that point an LLM would be useful by seeking connections in the database, and then highlighting overlooked patterns. The human interpretive performance would be banished from the field.
The irony is that archaeology, the study of old human systems, has not built a system for its own knowledge. If it ever does, the machines will be able to do what humans have already failed to: assemble the fragments into meaning.
Archaeology have cast themselves as the indispensable interpreters of fragments. If the field ever did the unglamorous work of standardising its data into a single coherent system, the need for interpretation would vanish. An AI could do the job faster and more consistently, leaving archaeologists as hunters and gatherers of data.
I realised this after watching an AI identify the location in an old photo I had. It broke the image into groups of pixel features, turned them into words, and matched them against reference material until the location appeared by deduction. Archaeology could be the same. If the fragments are properly described and standardised, the machine wouldn’t dream, it would just find.
I don’t mean to slag off solely at archeology. Much of the same dynamic has crept into all of science and just about all other academic subjects.
The incentive structure leans toward novelty, citation counts, and media traction rather than accuracy or reproducibility.
Logical thinkers get trapped between the careful limits of what the data can support, and the pressure to turn that into a “story” that grabs attention.
And because being wrong rarely carries consequences, the system accumulates bullshit.
So instead of building cumulative knowledge, we stack flimsy claims on top of one another, rewarding attention over accuracy.
Except in business where incompetence gets punished quickly because the measure is unforgiving: money. A bad model, a flawed assumption, or sloppy execution is measured and assessed immediately. It doesn’t guarantee wisdom, but it does enforce a minimum level of competence that academia and science don’t have.
Business systems optimise for throughput: turning forests into timber, fish into protein powder, oil into plastics and people’s attention into ad revenue. The waste, carbon, toxins, noise, distraction, isn’t priced at all, so it accumulates.
Add humans, we’ve only built one machine that is ruthlessly competent, and it’s ruthlessly competent at the one thing: converting nature into pollution.
So we’re good at one thing, slowly killing ourselves.
Maybe, just maybe, AI to the rescue? Just like it could save archaeology, it might save ourselves from ourselves. If AI is to rescue us, it will be by doing what we won’t: impose structure, cut through noise, and measure consequences without sentiment.
But it will take the will to let it take charge.
offshore westerly 