Supercomputer with Artificial Intelligence for Buddhist Education

[Indaviro]

We have seen the benefits of Buddhist resources being made available online in cyberspace. The Buddhist teachings have reached even beyond the extent of not only traditional Buddhist nations but also to places where there are hardly any form of Buddhist groups (lay or monastic). When I used to be a lay person, being active in social media, I have noted enough individuals who have no sangha within reasonable distances to where they live yet they can find ebooks, look up Dhamma talks and videos from home.

This may have been an idea that those who experienced in the Buddhist traditions have already thought about. Yet I did not see this specific topic being discussed and therefore I decided to post it, for two reasons. Firstly it’s to make aware those who may have the inclination for such interests. Secondary its to stimulate discussion around this topic to generate more views and opinions because it helps to appreciate the advantages of such an ‘imaginary’ or ‘hypothetical’ project as well as to be aware of the challenges related to that.

With the intention of facilitating Buddhist education I was pondering on the thought of different Buddhist groups from different Buddhist schools, traditions, denominations, and sects coming together to work on a joint project. It is to pool resources and build a supercomputer (ideally within a Buddhist library) with sufficient capacity (hard drive/solid state drive space, RAM and processors etc.) and have an artificial intelligence software installed on to this. Then to have all the canons, commentaries, sub-commentaries, works of important teachers of Buddhism (both current and historical), paracanonical texts, all transcripts of Buddhist talks / video talks from all teachers, transcripts of Q&As, Buddhist articles, texts on Buddhist history etc. Whatever the content that is available in Buddhism online can be installed on to this supercomputer. Whatever that is not available, where it can be digitised and then added onto the supercomputer. Permission can be obtained from the publisher and author or any party which hold copyrights so that textbooks can be included within the supercomputer. Basically all educational content in relation to all Buddhist schools can be included within the supercomputer.

The AI software is meant to analyse the information and made to go through cycles through the content within the supercomputer to make itself familiar. Several Buddhist dictionaries, glossaries and encyclopaedias can be installed so that the AI becomes familiar with the Buddhist terminology so that it can interpret the content through a Buddhist lens and narrow down possible interpretations. Common choices for translating words in scriptural languages can be made available so that when it comes across translations (English translations) then it will be able to refer to the meaning of the terms from the original scriptural languages (Pali, Sanskrit, classical Tibetan, traditional Chinese etc.). As part of ‘beta testing’ a series of common questions can be presented to the supercomputer with AI where it will be given enough time to plough through all the texts and find a complete answer.

A website can be made where the output of the supercomputer is made available to. Users can access the website and ask questions, or ask it to write essays and the AI will give its answer. The user will have the option to select the type of answer based on which school or tradition of Buddhism. For example there can be a panel with Theravada, Mahayana and Vajrayana and the user can have the choice of just selecting Theravada. And then within Theravada one may select Thai tradition or Burmese tradition. Within Theravada one may have the option of choosing Suttas only and omitting answers from commentaries and sub-commentaries. The user may also go for an overall answer where the AI can present the answer in stratified form into different headings and sub headings (such as Theravada and then Theravada Sutta-based and Theravada commentarial etc as well as Mahayana - Pureland, Mahayana - Soto Zen, Mahayana - Rinzai Zen etc.). In addition to the stratified answer based on texts, schools and traditions, it will also write an overall ‘synopsis’ like answer with its own conclusion at the bottom.

For example, lets that a user asked the question on ‘please explain the practice of the ‘Buddho’ mantra’. The AI will look into the scriptures, commentaries as well as the content from Buddhist teachers to blend it all in to come up with an answer. If a user wanted to selected ‘Theravada-Thai tradition-forest tradition’ then the AI will still be able to look through the entire of the content within it but present only what is relevant, where that answer is based on the Pali canon, commentaries as well as based on the works of Thai forest tradition masters. The user may tick other Buddhist traditions and the AI will present answers based on those traditions of choice.

The AI should be able to perform more complex tasks such as write essays and even academic research papers. It should be designed to facilitate the process of academic study in Buddhism where a Buddhist academic can provide the query or specific topic and then the AI can make a list of resources as well as generate its own thesis which will be very useful for universities and colleges where Buddhist studies are taught (as long as it used in an ethical manner). The AI can be used for comparative Buddhist studies where cross comparison can be made between Buddhist traditions. For example, one can pose the question - “what are the similarities and differences between Shikantaza and Dzogchen and how they both evolved” and then the AI can look through all relevant scriptures, look up textbooks and talks from well known master practices in those practices to come up with an overall answer with topics, sub topics, chapters etc.

The ways to facilitate Buddhist education have changed through the centuries. They are all useful and none are obsolete. The writing down of the Pali canon as well as the other Buddhist canons played a big part in Buddhist education. With the invention of printing technologies, Buddhist teachings were made available as textbooks. Then later on with the advent of the World Wide Web, the Buddhist teachings in the form of ebooks, online articles, audio and video talks are available to all who have access to internet. The supercomputer equipped with AI and Buddhist texts with all Buddhist resources will be the next milestone to cross for the sake of further advancement of Buddhist education for the benefit of those who are interested to deepen their knowledge and practice in Buddhism in almost all different orders of magnitude.

[NgXinZhao]

It’s been done, although I dunno to what extend the data is fed to it like you asked.

Training AI models on the suttas

There’s a generic Buddhism one as announced here NORBU the Buddhist chat-bot and this is a special EBT one, as the problem with generic Buddhism is as we know, on the deeper topics, views diverges. Feel free to explore. Haha just try to ask the 2 different versions this question: jhanas do not require 5 physical senses to shut down is that true?

There’s the EBT version (EBT nothing lite) and Generic version (Something deep). See this post to understand the things in the bracket. Proposed EBT naming nomenclature

You can ask the creators in facebook group. Redirecting... to add in feedback and whatever you want them to add in etc.

Also, take note of these series of articles against AI: AI-1: Let’s Make SuttaCentral 100% AI-free Forever

Which goes until 16: AI-16: What to do?

If you wish to reply to that, do reply on the AI-16 one.

[Malunkyaputta]

I am very skeptical about AI and philosophical/doctrinal/metaphysical questions.

When ChatGPT came out, I tested it numerous times on Kantian philosophy.

It regularly came up with nonsense. When drawn to its attention, it acknowledged it and came up with even more nonsense.

The eternal questions of mankind are not so quickly going to be resolved by AI.

[NgXinZhao]

Indeed, I just asked the chatGP4 and our norbu AI bots to list 227 rules for monks, and they cannot do it accurately.

[Indaviro]

Thank you Venerable. I see the Buddhist world has made a start already. It’s good to hear that it has started for EBT. Later different groups may put a combined effort to include content from different schools. Ideally from all traditional schools of Buddhism in all languages.

[Indaviro]

It’s always good to have a healthy form of skepticism. Given that AI is still in its sapling stages we can’t expect a lot from it. However it has the potential to grow into something more mainstream with time (probably sooner than we think). ChatGPT is one example of an AI software.

For Buddhist educational purposes one can ‘train’ the AI to give more reliable responses. With anything, not just AI it needs to go through a phase where it is tried and tested before it is released into the public. AS for Chat GPT it is not specifically designed to address questions related to Kantian philosophy. Softwares can be designed to deal with topics and contents of a more specific nature. Either the AI can learn and be familiar with the content or else the recognition of Buddhist principles and themes can be inbuilt into the neural network of the AI.

[sujato]

AI is inherently unreliable and unsuited for Buddhist scriptures. All it does is produce streams of data into which humans hallucinate meaning.

What’s holding Buddhism back is lack of faith in Dhamma and in humanity. Buddhists have lost confidence in our own capacity to accomplish great things. AI feeds off of that. If the underlying theme of advertising is that “you need this, because you as an individual human are inadequate”, then the theme of AI is, “you need this, because you as a human race are inadequate”.

Speaking personally, and as someone who contributes to Buddhist learning and education on a daily basis, I sincerely and deeply pray that I never have to read a single sentence of machine-generated text ever again for the rest of my life.

[Dogen]

@Malunkyaputta

The eternal questions of mankind are not so quickly going to be resolved by AI.

The answer is 42.

[Dogen]

@sujato

What’s holding Buddhism back is lack of faith in Dhamma and in humanity. Buddhists have lost confidence in our own capacity to accomplish great things. AI feeds off of that.

Do you feel SCVoice and BuddhaNexus also fall under that definition, Bhante?

[sujato]

@Dogen

SCVoice

Voice uses artificially generated text-to-speech, which is a very limited and specific use of AI within highly defined constraints, and achieves accessibility for people who cannot otherwise learn.

As for BuddhaNexus, I’m kind of on the fence. It basically maps connections, or supposed connections, so it points you in a certain direction so you can look at things. In that way it’s like a search engine or product recommendation algorithm. I personally haven’t found it useful, but I also don’t think it’s harmful if used properly.

In neither of these cases are you taking knowledge, chewing it up, and spitting it out as something new.

[Pasanna]

Your motivation for this project seems to be to support and create community for lay buddhists who are in areas remote from physical community.

I don’t see how having a database and a robot to talk to is going to create that sense of healthy discussion and community which can be so beneficial for the development of dhamma practice. There are some wonderful Discord (and other communities) which meet regularly for dhamma discussion and sutta discussion specifically. I feel like these kinds of online networks are far more useful than a database of text which are shuffled, stirred and churned out by AI. What online Buddhists are missing out on is connection, community and ritual. They are already listening to too many dhamma talks and reading too many dhamma books.

A better solution to the problem of lack of community is to create community or a mechanism for people to create community. Real human to human, cups of tea and smiles community.

[Indaviro]

There is something which I forgot to mentioned directly in the original post. In recent years we have seen a very rapid boom of AI however it is still in its sapling stage. Just like all new inventions that we have seen in the past century, they all needed to be given time to develop as well our human minds to mature around the way of those new inventions. Likewise I am hopeful that since its beginning nothing has been absolutely perfect and they all need to be given time and space to develop and further evolve.

A specific advantage I forgot to mention originally - lets say someone who had interest to learn everything what the Buddhist texts as well as Buddhist teachers had to say about Dependent Origination / Co-dependent Arising (Paticcasammuppada). Ideally one would have to go through all the content of the Buddhist scriptures and provided they are interested what the commentaries have to say, then she would also have to read those commentaries. Following which she would have to read or listen to all transcripts or audio/video files on those teachings. Trying to search for all of this is a big enough task already let alone having to go through all of them.

Now I believe an AI software that is designed to handle this task can fast forward this process where it will save time as well as give the complete information (or near complete) information on the Paticcasamuppada. It will search and come up with any sutta or any kind of scriptural passage or stanza which mentions the Paticcasamuppada. Even if the term is not explicitly mentioned, and instead if that is explained in description only, the AI may have markers to recognise and associate such descriptions with the topic of interest and also make reference to that. Then it will provide its own description of the Paticcasamuppada. It could be even designed to update Wikipedia articles to make them more consistent with Buddhist texts as well as teachings from Buddhist masters.

But let’s say that one is not interested in what the AI has to say. At least the user can look up the sources that the AI has come up with and directly refer to them (which is pretty much already there in search engines). However the advantage of having a supercomputer with Buddhist content is that it now has access to just Buddhist content (ideally from all schools of Buddhism, as well commentaries, sub-commentaries and works, articles, textbooks of Buddhist teachers).

There is already evidence to show that AI is better at performing tasks than humans. For example, in its initial stages, AI was not able to recognise faces and identify them. Now, a particular narrow AI has the ability to identify faces much better than the average human being and a lot faster. We need not see this as human inadequacy. Rather than that, we can see that it was human beings who gave birth to AI, and therefore this just one of the ways we can appreciate that human beings are not only capable of understanding Buddhism but also designed an AI bot which can present the Buddhist information much more efficiently and more accurately which can happen with time with further development of AI software.

I can admit at present, AI is going through a phase where it is still making mistakes, so with further advancement of those technologies, it will get better and reach the point where it will surpass the ability of humans to gather information and analyse it as well as make its conclusions. We already have non-AI software to replace certain human functions, but this is seen as facilitating certain processes to make things more easier and efficient. We already have non-AI information databases to facilitate searches and come up with sources for topics of one’s choice. AI can be seen as a further extension of this same process.

Now I can appreciate that one may think that there is also the purpose of AI to replace human interaction and appear like a ‘friend’. Even though that is seen as a possibility by some, that is not what I was suggesting in this post. Of course I think having human interaction with real humans is a great thing. We need more heart felt connection with human beings. I don’t think AI as applied to Buddhist education will be an obstacle to that.

[Indaviro]

There was something very interesting that came to my mind about 2 weeks ago whiles I was in a different monastery. One of the best ideas I ve thought of. It’s a shame, given that had it comes across my mind earlier I would have included that in the original post. With time, a lot of different ideas may flood into my mind regarding this topic so I hope to write an essay or dissertation on Buddhism and AI - its use in Buddhist education. Here is what arose in my mind.

In the Buddhist world there are Buddhist scholars who are from faith oriented backgrounds (traditionalists) as well as those who are outside of that. The AI which I have in mind should be able to identify from their works (thesis, articles etc) how they arrived at conclusions based on scripture, commentaries and other important texts as well as what is happening in the Buddhist world regarding practice. The AI should be able to draw logical pathways between the sources (text or practice) and then link it to the conclusions. When the AI is exposed to the works of authors with enough examples as part of their works, the AI software should be able to identify the patterns of their logical pathways of thought and analysis that was used to help the authors come to conclusions based on what source material that they used.

For example, if we take Analayo Bhikkhu, the AI Supercomputer may be fed all of his works. Then the AI Supercomputer will look through all of that material, and then note the references, and then identify (or reverse engineer) his logical pathways of thought which made him arrive at the conclusions that he arrived at where the supercomputer AI not only thinks but also is able to reason the same way as Bhikkhu Analayo. This can be done for Buddhist teachers such as Thanissaro Bhikkhu, Sujato Bhikkhu, Bodhi Bhikkhu etc as well as scholars/academics who are not from faith oriented backgrounds such as Peter Harvey, Richard Gombrich *etc. Then these same logical pathways of thought to derive conclusions, once it is fully identified by the AI Supercomputer, then it can apply at to other parts of the Buddhist texts which may not have been covered by such scholars to for the Supercomputer AI to author its own materials. For example, Bhikkhu Analayo wrote a thesis on the Satipatthana. The Supercomputer AI can use a combination of logical pathways of different scholars and academics and author a thesis on Satipatthana which combines all of their thinking and analytical strength. But not only that but also apply that same combined logical pathways of thought to author works on other aspects of the scripture, commentaries and practice.

If a university or college (like SOAS) can type in its academic query and add more description on what it exactly needs and the Supercomputer AI can come up with a thesis or articles and it will help complete the works and projects that are required by universities and colleges much more efficiently and a lot quicker and on top of that it will be the intellectual strength of multiple academic, scholars and practitioners. The range of works that it can generate regardless of the broadness or specificity of the topic would be almost limitless (eg. a thesis on Paticcasamuppada or a specific query such as “the development of the divergence of the doctrines within the Abhidharma”). The responses will not only be limited to orthodoxy but be inclusive of orthopraxy as well.

In summary to sum it up - for the AI software to identify the thought patterns on scholars, teachers etc and then combine that power and apply that to any kind of Buddhist doctrine or practice.

• This is not to say that they lack faith or commitment to Buddhism, but instead it is to highlight the fact that their main commitment towards the Buddhist knowledge system is based on academia rather than faith (at least on the external level).

[Dogen]

I will be eternally baffled with our Buddhist tradition trying to find anything other than conditions in their present existential scheme, and why those conditions can’t be repeated elsewhere (i.e “Machines can never be humans”).

Xenophobia, Human arrogance, Atman views creep into our discussions about AI abundantly.