I’ve become a little bit obsessed with AI. Like many, I played with early versions of ChapGPT. I found it curious, but was pleased to discover that it didn’t know what I knew. And it sometimes just made stuff up! So I retreated to the safety of more reliable digital resources, such as Sergey Minov’s Syriac Bibliography, Beth Mardutho’s lexical and corpus tools, and Syriaca.org.
Things changed just recently, however, with the launch of Oromoyo.ai. I had already been using DeepL, which is a brilliant AI assisted translation tool for a variety of modern languages. I had even played with Google Translate for Latin. So when I heard about Oromoyo.ai, I wanted to try it out immediately (more below). This led me to go back and look at developments in AI and how it might impact my work.
There is a lot of information out there on AI! And it’s changing constantly. I have found Ethan Mollick to be a particularly helpful guide. He is a professor at Wharton, where he co-directs the Generative AI Lab. He researches the use of AI in business, and has produced some really useful resources, such as the five part video series introducing AI for teachers and students, done with his wife, Lilach Mollick, also a professor at Wharton; an information packed Substack, One Useful Thing, and a new book published last year, Co-Intelligence: Living and Working with AI. Mollick’s general advice is to just start using AI tools and see what works! But he also has some specific advice too.
The secret to AI is that it is trained on vast amounts of data (Large Language Models), and can generate new material based on this training and the neural network architecture that it uses. This means that AI knows a lot and can do many different kinds of things. One of the other important things to understand about AI is that it is improving rapidly, moving from high school, to college, to graduate level knowledge and ability in just a few years. So, whatever you think about these tools now, know that they are just going to get better.
So, now for the four tools that I’ve found to be most useful so far:
Oromoyo.ai
This tool translates text and speech from and into Syriac. I have only used the Classical Syriac to English translation option. There are other features too, designed to help learners. “Founded in 2024 by Benjamin Faal, oromoyo.ai leverages the power of artificial intelligence to make Aramaic language learning accessible, engaging, and tailored to the needs of today’s users.”
As for Oromoyo.ai, I would recommend just trying it out using whatever text you are working on. Predictably, it seems to be better with genres and authors for which there is more training data. You are limited to 500 characters, which is the main limitation–context matters for generative AI, so I would imagine that translating larger amounts of text might make the tool even more accurate.
This tool is absolutely not a substitute for expertise or reading fluency. It does not yet produce the quality of translations that DeepL produces, for example, but it is only a matter of time. And what it can do now is shockingly impressive.
USES: Drafts, quick translations during research, ideas for dealing with tricky passages.
Gemini–ChatGPT–Claude–Perplexity
AI Chatbots know stuff. They still hallucinate. That has and will continue to get better, but might not be completely eradicated. But it shouldn’t stop us using them for all kinds of things. Find a chatbot that you like and start thinking about how it could be integrated into your research, learning and teaching. This will often require thinking about how one thinks and how we work and teach. If you want some ideas, just ask the chatbot.
One piece of advice from Ethan Mollick: always use the Frontier Model Chatbots, which are built on the most cutting edge AI models, and always upgrade to the best version.
USES: Summarize information; teach you stuff tangential to your field; analyze texts; produce drafts of abstracts; outline papers and classes; etc.
Elicit
I have just started to use this tool, but it is the closest thing that I’ve found to an AI research assistant for academics (here’s an introductory video). The three options at the moment are 1. Find Papers; 2. Extract data from PDFs; 3. List concepts. I have played a little with the Find Papers option. This is oriented around published research and every source that it provides has a DOI and is linked to a published paper. Questions you ask will elicit a list of research paper and books that you can then interact with in various ways. So, this is more than search, it quickly become research. Each query is saved as a Notebook, that you can refer to at a later stage or delete if no longer needed. I am eager to use this more and think it has the potential to be a really interesting research assist tool.
USES: Research in related and ancillary fields; curiosity; summarizing content; orientation to a new field.
NotebookLM
This is my favorite research tool. I use it to create “Notebooks” for research questions that I have. A notebook is essentially a research folder that contains up to 50 items. All of mine are filled with PDFs of books and articles. But they can also be other text or audio files, links to Google docs, websites and YouTube videos (imports the text transcript). Once you have populated your notebook, then just start asking it questions. I have recently created a notebook on Text Criticism and Editorial Technique, for example, and it has been splendidly instructive and quite fun to engage with these sources through this means. This is by no means a substitute to deep reading, but I have found it to be a useful adjunct and perhaps palate cleanser for deeper reading and research.
This tool also forces me to think more carefully about my own research process and learning style. Mollick argues that teachers are often the best at using AI tools because you teachers are great at breaking projects up into stages and have the students work iteratively through them. This kind of thinking is also useful with AI research tools. There is a way of thinking and asking questions that helps get the most out of the tools. This is called Prompt-Engineering. But Mollick argues that the tools are increasingly become sophisticated enough that they will actually prompt the user to get the most out of them!
Studia Syriaca 