Hello! My name is Chandupa. I am the representative of a team that recently earned an acceptance into the Northeast I-Corps Regional Program: Princeton University I-Corps Propelus for the Spring 2024 cohort. I-Corps is a National Science Foundation (NSF)-funded immersive program designed to help better understand if and how technology can meet unmet needs.
We wish to understand people’s experiences trying to learn about Buddhism in order to create technological solutions to solve problems that people face in this process.
I cordially invite you to an interview to share your experiences with studying and practicing Buddhism!
I-Corps requires these interviews be done live (ranked in order of their preference):
in person
video call
audio call
If you are unwilling or unable to do a live interview, we can also collect information through a questionnaire, but it would not fulfill the following program requirements, which are:
March 25th to 27th - complete 3 interviews
March 27th to April 17th - complete 20 interviews
If you are interested, please PM with the dates and times that you are available.
Also, please feel free to extend this invitation to anyone else who might want to share their experiences with Buddhism in an interview!
Two of us are currently located in NJ, USA and one of us is in Thailand, but we have been awarded funds to travel to attend events, participate in conferences, and join professional associations through which we can interview people. Any recommendations would be helpful information.
Please PM me to inquire further or schedule an interview.
We are very eager and excited to hear from you about this topic!
Thank you in advance for your time, consideration, and input.
For me, I remember the teachings in my own paraphrased phrases in my mind, so when I want to search for a sutta to quote, the search engines are not so helpful as I don’t remember the exact phrase the translators uses. Like snakes can be viper, serpent, etc. I rely a lot on forums to find sutta citation fast. If the AI chatbot can include all forums materials as well as thesaurus search for all keywords, it would be nice.
I think for LLM one would of course use the public discourse but I imagine the model will be quite bad and unreliable in making things up and otherwise mimicking the same confusing talking points.
The amount of analogical reasoning that goes into making sense of the texts is ridiculously challenging even to humans and i wouldn’t hold my breath waiting for an AI capable of anything close to what a human can do in forming complex arguments and evaluating merit where non-classical logic is utilized.
I think AI will become a substitute for much of repeat beginner’s questions and aid in translation & vocalization but that is probably the extent of it’s utility.
A custom trained model would be interesting, just to see how advanced & reliable it can get, but it will probably never be made entirely safe to use because it is essentially just making things up which happen to mimick real rational answers but it will never give real rational answers based on merit evaluation like a human does.
There are not many people now who have the type & the amount of content that one would need to train a model.
The best model would be of a single person or a group explaining a single interpretation of the texts for many years. I wonder if it is humanly possible for one person to write enough about this, how many pages of text are we talking about? Might need millions of pages or something like this? Nobody can do this if it is so much, 3000 pages is the best aggregate that i know of in existence, of a single person & a single interpretation.
So one could modify one trained on the public discourse and have it mimick a particular person where it can. This custom model i think could be better than a public model but i wouldn’t feel safe in advising people to use it because it can make up an answer to anything even where the person would say ‘I don’t know, let me research this and get back to you when i have an answer’.
Another fundamental issue is that one would want to edit & moderate the content on which it is trained and things it says. The public discourse is also moderated, like this forum has ToS and policy which makes people self-censor and this will be a deficiency in any model that one makes based on what people say here.
A LLM is essentially akin to listening to a foreign language enough to participate in discourse without any idea of what any of it means. It’s impressive but it can never be anything more than that.
Yes. It’s really frustrating to see people continually call LLMs artificial intelligence. There is no intelligence. Just statistical prediction. They are not even large knowledge models.
I’m all for using technology to help people. But these tools are not appropriate for this use case.
