New AI tools advance ancient text reconstruction and brain image decoding

Researchers have introduced two significant artificial intelligence developments: Apollo Restore and BrainIt.

Apollo Restore, developed by researchers from the Austrian Academy of Sciences in collaboration with Mistral AI, SAIL Reply, and Northeastern University, is designed to reconstruct ancient Greek texts. Trained on 600 million words with 24 billion parameters, the system can propose words or passages to fill gaps in damaged papyri and inscriptions without requiring the user to specify the exact length of the missing fragments.

Separately, the Weizmann Institute of Science in Israel has developed BrainIt, a bidirectional AI system capable of reconstructing images from neural activity in real time. Unlike traditional scanners that require dozens of hours of training, BrainIt can adapt to a new subject after only one hour of individual recording. The technology uses neural networks to generate detailed visual representations from captured brain activity, helping scientists understand how the nervous system processes information.