Deep neural networks are increasingly used among different industries. However, apart from popular domains like finance or marketing, the usage of artificial intelligence can also greatly help for example historians. And now they do have their own tool.
Ithaca
In general we do not think that ancient Greek inscriptions would be in any kind of interesting area of artificial intelligence usage. On contrary, the field of epigraphy seems to be the domain of historians and other scholars. Now there seems to be a complete change of situation in this industry. The tool, which is called “Ithaca”, is made by Google DeepMind. It can successfully perform a range of activities. The main task of the system is text restoration. What is more, identifying the geographic location and time with the date of the construction are definitely game-changing for historians. The number of possibilities to gain new knowledge from this tool is extraordinary. Ithaca is based on a previous tool which was called Pythia. The new system accuracy is 62 percent compared to 25 percent for people working alone. The accuracy ranges even up to 72 percent when the tool and scholars work together. Finally, it is said that Ithaca can be easily modified to work with different ancient scripts, making it a versatile system.
The process of making
The first tool – Pythia, was trained on an enormous dataset of 35 000 inscriptions. In addition, there were used 3 million words from that period of history. Pythia in general was suggesting the possible option of text understanding and filling. It was very much relying on cooperation with the historians. The Ithaca presents higher accuracy and additional features of the suggested time and place from which the text comes. It was trained on more than 78 000 inscriptions combined with descriptions about origin date and location.
Implications
Overall potential of Ithaca in ancient text restoration is very high. This process done only by human is time and effort consuming, now it can be changed. Furthermore, the information about when and where the text was written is crucial to expand the current state of knowledge rapidly. For now, it is not clear how fast it will be adopted and used across different teams of scholars, but the tool is promising. It could greatly speed up the research about the ancient periods of history. What is more, if it would be truly versatile many regions of the world could also learn more about their history. The system is available online for everyone to try. In the current state, the best usage of Ithaca is the close cooperation with professional historians. Lastly Ithaca should continue to be developed for higher accuracy and new features. It seems that even not obvious industries can benefit a lot from the artificial intelligence tools offered by today’s technology.
References
https://www.wired.com/story/deepmind-ai-restore-ancient-texts/
https://www.ox.ac.uk/news/arts-blog/restoring-ancient-greek-inscriptions-using-ai-deep-learning
The idea that artificial intelligence can serve historians is lunacy. It is a wonderful meeting of the past and the future. But it also poses a fundamental challenge. This innovation leads us to think that potentially, in the long term, artificial intelligence could replace the work of historians.