In previous installments of this series, we looked at scholarly and popular publishing from Buddhist book publishers. In this installment, we’re going to explore online open source, not-for-profit, canonical, and liturgical works, and particularly how artificial intelligence is transforming workflows in these endeavours.
First, a few vignettes from the leading edge . . .
The Khyentse Vision Project (KVP)
The KVP has as its mission to preserve, translate, and digitize the collected works of Jamgon Khyentse Wangpo (1820–92), who was the first in the Khyentse lineage and one of the founders of the Rimé school. The collection spans 47 volumes comprising more than 30,000 Tibetan pages of text. So far, KVP has published several hundred texts, freely accessible in their online reading room. The organization estimates it will take another 10 years to complete the project. They are exploring AI to expedite the process.
MITRA: Machine Translation for the Languages of the Dharma
MITRA is an AI-powered tool for automated translation, semantic search, and philology, using a Gemini API. The developers’ goal is harnessing AI technologies to promote the scholarly study and personal practice of the Dharma, and to accelerate academic and individual research through open-source collaboration on datasets, models, and applications. MITRA supports Sanskrit, Tibetan, Buddhist scriptural Chinese, and English. It is a collaboration of the Tsadra Foundation, the Kumarajiva project (a project of Khyentse Foundation), AI4Bharat, Monlam.AI, and IIT Kharagpur.
Imagining a Buddhist AI
In October 2024, UC Berkeley hosted a conference titled AI and the Future of Buddhist Studies. Panels included “Uses of Advanced Computational Methods for the Analysis of Buddhist Texts; Creation and Analysis of Digital Corpora,” “Machine Translation of Buddhist Texts,” and “The Impact of AI on the Training and Work of Buddhist Scholars.”
The same month, SOAS University in London hosted a similar conference on developments in Buddhist AI, sponsored by Khyentse Foundation. One of the presenters, Ven. Miao Guang from Fo Guang Shan, Taiwan, gave a brief overview of their work in the field of digitizing and translating Buddhist texts using AI, summarized below.
In the first phase (1970s), Prof. Lewis Lancaster at UC Berkeley began a project to digitize the 84,000 woodblocks in the Tripitaka Koreana. At the time, given the limitations of the digitization process, it was a daunting task.
Digitization of the Chinese Mahayana canon began in the 1990s. The CBETA Taishō Set was digitized in the 1990s.
In the early 2000s, Fo Guang scholars began digitizing the Fo Guang corpus, which comprises 2,000 volumes and covers 16 canons. So far, out of about 400 volumes of the Fo Guang canon, their project has completed about eight volumes. Out of 45 volumes of other Chinese Buddhist canons, their project has completed three.
Like the MITRA project, Fo Guang Shan is slowly building a data set for translation, semantic search, and philology. Ven. Miao predicted the project would take 20–30 years to complete.
The limitations of Buddhist AI chatbots
Ven. Miao explained to the conference that it takes between 10 and 100 gigabytes of information to train a large language model (LLM) to converse effectively. The collected teachings of individual Buddhist masters are themselves too small to properly train an LLM, since they would typically only comprise 1–100 megabytes. AI engines can parse nine billion tokens per second,* so attempting to have an AI chatbot creatively “teach” the Dharma based on the teachings of one teacher, or a select few, results in outputs that are repetitive and overly generalized because they are trained on limited data sets. However, this does not mean folks are trying to make Buddhist AI chatbots. Following are some examples.
Roshibot is a Zen AI chatbot trained on the collected works of Shunryu Suzuki Roshi. Pseudo Shinran is a Jodo Shinshu AI chatbot. Norbu AI is yet another Buddhist chatbot, from Malaysia. There are also Buddhist chatbots such as Mindar, Sati-AI, BuddhaBot, and so on, with more being developed all the time, usually as experimental forays into the world of artificial intelligence.
While these efforts may be disappointing so far, it’s worth remembering that the artificial intelligence being brought to the task is the most incompetent AI we will ever see. It’s only getting more sophisticated, albeit in fits and starts.
Buddhism and the case against artificial intelligence
As commendable as the above ventures in publishing canonical Buddhist works are, and as interesting as the learning going on in the development of Buddhist AI chatbots is, I am not hearing very much at all from Buddhist teachers about the dark side of artificial intelligence. Rather, everyone seems to be enthralled by the technology in spite of the deep contradictions surrounding Buddhist practice in an AI-powered future.
As dazzling as is the simulacrum of wisdom displayed by AI, it is entirely antithetical to Buddhist compassion, in the larger context of prosocial contributions to society. It is also antithetical to the importance of realized lineage holders who teach the Dharma with the authority of actual awakening rather than glib verbiage and linguistic probability models from machine learning. The notion that one can dissociate Buddhist wisdom from compassionate engagement in alleviating suffering is a specious argument that unfortunately has been a recurrent theme in the evolution of Western “modernist” Buddhist discourse.
There is no shortage of books, articles, and presentations about the dangers of artificial intelligence. Consider, for instance:
• The insatiable demands for electricity and water by data centers
• The hollowing out of critical thinking in education
• The vaporization of viable career paths for graduates replaced by AI
• Unelected tech billionaires in control and increasing wealth inequity
• The erosion of privacy in a surveillance panopticon
• The erosion of democracy in a world of deepfakes
• The weaponization of AI in geopolitical struggles
• The inability of legal guardrails to protect individuals
• Intensifying suffering and delusion
In conclusion
In an interview with Mariana Restrepo in Lion’s Roar, February 2025 , KVP founder and director Dolma Gunther mentioned in one sentence that they tried to be carbon neutral in their efforts. That’s commendable. However, weighing the benefit of that effort against the overwhelming negative consequences of our growing reliance on machine learning and its progeny, I believe that it would be much more in keeping with the Buddha’s ethical instructions if we were loudly speaking out against artificial intelligence at every opportunity.
Furthermore, I have grave doubts about the value of continuing to translate canonical texts as a boon in and of itself. Don’t we already have lots of Buddhist instructions? Isn’t it time we put them into practice? I know that the process of bringing the Dharma to new civilizations over centuries has allowed me the ability to practice and has involved immense sacrifice by thousands of earnest translators. But to some degree, I feel these efforts can veer into extremes of scholasticism, akin to figuring out how many angels can dance on the head of a pin or the correct pronunciation of “shibboleth.” That said, I’m certainly open to being persuaded otherwise. Full and frank discussion is more important than ever!
For a much more in-depth overview of how artificial intelligence is transforming the world of publishing, I strongly recommend The AI Revolution in Book Publishing: A Concise Guide to Navigating Artificial Intelligence for Writers and Publishers by Thad McIlroy.
Coming up
I hope you have found this series on Buddhist publishing informative and eye-opening. In the summing up installment, my goal is to bring you further thoughts from a variety of stakeholders in the field, responding to the perspectives raised in the first five installments. See you then!
* A token can be a letter, a number, a punctuation mark, or some other typographic element.
See more
Khyentse Vision Project
Dharma Mitra
Tsadra Foundation
The Kumarajiva Project
Building AI for India!
Monlam.AI
Indian Institute of Technology Kharagpur
AI and the Future of Buddhist Studies (UC Berkeley)
Imagine a Buddhist AI… (YouTube)
Techniques for Collating Multiple Text Versions in the Digitization of Classical Texts (Chung-Hwa Institute of Buddhist Studies)
Roshibot (Shores of Zen)
Pseudo Shinran – AI-powered Shinran Wisdom (YesChat AI)
Norbu AI
The AI Revolution in Book Publishing: A Concise Guide to Navigating Artificial Intelligence for Writers and Publishers (Leanpub)
Exploring AI and Buddhism: A Conversation with Khyentse Vision Project (YouTube)
Related features from BDG
So, You Want to be a Buddhist Author, Eh? Part One of A Six-Part Exploration Into the World of Buddhist Publishing
Buddhistdoor View: The Dharma Futurist’s Hope for AI in a World of Suffering
In a World of Human Ignorance, Can Artificial Intelligence Help?
The Rise of Artificial Intelligence and What it Means for Our Jobs
You must not be looking very hard. Bhante Sujato has written on this extensively on discourse.suttaCentral.net.
If you’re looking for Buddhist text translations, it’s impossible to miss Sutta Central. Sutta Central founder Venerable Sujato has written extensively on AI translations of Buddhist texts and there are numerous massive discussions about AI generated Buddhist texts on Sutta Central’s forum https://discourse.suttacentral.net/ . AI translations of Buddhist texts, AI generated posts and AI generated images are not allowed on the forum.
I appreciate your views very much John. Thank you. Humans, and human translators, are not replaceable. Practice is essential.
I found Geshe Tenzin Namdak’s note about living in a material world really hits when he ties ancient ethics to AI today. That emphasis on balancing tech with compassion reminds me of late-night debates in the monastery about how gadgets shape inner life.
I’m struck by the question of whether AI has Buddha-nature, and the line about machine learning mimicking compassion makes me wonder where genuine intention begins. Reading that nuance, I recall how memory-based writing can feel almost mindful in its repetition, yet still lacks true awakening.
I found the line about AI challenges in Buddhist authorship really resonant—artistic struggle mirrors the inner discipline of practice. Reading that made me think of how I once tried to write during a late-night gaming session and saw how distractions steal quiet, mindful moments.