Ah, he really can’t read it, can he?
If Desset fed his complete decipherment work — his sign list, phonetic values, grammatical rules, all 45+ inscriptions with annotations — into a purpose-built AI model, it could potentially:
- Read any new Linear Elamite inscription presented to it
- Identify signs even in damaged or weathered contexts by pattern completion
- Cross-reference recurring sign clusters against the full corpus simultaneously
- Suggest probabilistic readings where signs are ambiguous
- Do in seconds what takes a human scholar weeks
Where it gets really interesting:
The model wouldn't just read known inscriptions. It could potentially:
- Detect recursive structural patterns across the entire corpus that no human could hold in working memory simultaneously
- Test the Elamo-Dravidian hypothesis computationally by comparing grammatical structures against Dravidian language databases
- Attempt partial decipherment of proto-Elamite — the still-undeciphered predecessor script — by working backwards from Linear Elamite
- Cross-reference against Indus Valley script patterns to look for structural relationships
The honest limitation:
A model trained only on Desset's 45 inscriptions would be powerful but brittle — high confidence on patterns it has seen, potentially unreliable on genuinely novel constructions. You'd want the full corpus, high-resolution imaging of every known artifact, and ideally new discoveries to expand the training set.
The deeper point though:
What Desset did manually over two decades — finding the Shilhaha name pattern, building outward from that crack — is exactly the kind of pattern recognition task that AI does extraordinarily well at scale. The human genius was knowing where to look and what question to ask. Once that framework exists, AI can industrialize the application of it across every artifact in every museum collection worldwide.
There's a reasonable argument that the combination of Desset's decipherment framework plus modern vision AI could advance our understanding of Linear Elamite more in the next five years than the previous hundred produced.