AI in Ancient Language Decipherment Quiz

Answers and explanations appear as you take the quiz.

1. What does DeepMind's Ithaca model do for ancient Greek inscriptions?
   • It restores missing text and estimates date and place of origin
   • It carves new inscriptions in stone
   • It only counts the number of letters
   • It translates them into spoken audio
2. How do sequence models help when an ancient text is broken or worn away?
   • They physically repair the stone tablet
   • They predict the most probable missing characters from surrounding context
   • They delete the damaged section
   • They translate only complete sentences
3. What character-alignment assumption does the 2019 neural decipherment paper encourage for cognates?
   • Every word has exactly the same length
   • All characters must be translated into English
   • An order-preserving correspondence, because character reorderings are assumed to be uncommon
   • The order of every character must be reversed
4. What is cross-lingual alignment in the context of decipherment?
   • Aligning the physical fragments of a tablet
   • Translating one modern language into another
   • Lining up letters to make them look neat
   • Mapping an unknown script's patterns onto a known related language to test translations
5. How do these models typically represent a script internally?
   • As sequences of tokens whose ordering probabilities are learned
   • As a single image with no structure
   • As random numbers with no relationship
   • As musical notes
6. What role do human scholars play when AI proposes a decipherment?
   • None; the AI's output is automatically final
   • They test the AI's hypotheses against historical and contextual evidence
   • They only retype the AI's output
   • They translate the AI into Latin
7. Why does the 2019 decipherment model use a minimum-cost-flow formulation?
   • To repair physical damage to stone tablets
   • To measure the age of ink
   • To constrain the overall matching of words across the two vocabularies
   • To generate an unlimited number of unrelated translations
8. What input does the 2019 decipherment experiment use from a known related language?
   • A bilingual translation for every ancient sentence
   • Only photographs with no text information
   • A verified translation key for every possible symbol
   • A non-parallel corpus that supplies linguistic correspondences to test