Compassion at the Frontier
Chris Scammell's reflections from the 2025 Mind & Life Dialogue on AI
I’ve just returned from a ten-day trip to Dharamsala India, where I had the pleasure of attending the 39th Mind & Life Dialogue: “Minds, Artificial Intelligence, and Ethics.”
For nearly forty years, Mind & Life has hosted conversations at the intersection of science and spirituality alongside His Holiness the Dalai Lama. This year, Artificial Intelligence was the centerpiece, approached from a few different angles:
Minds - What is intelligence? What distinguishes living minds from artificial minds?
(Dis)entanglement and meaningful relationships - How are human minds intimately entangled with artificial minds?
Collective narratives and possible futures - How do the stories we tell about AI create self-fulfilling prophecies?
Diversity and ethics - How might a more complex understanding of intelligence impact the AI systems we develop and the regulations we put in place?
Education - How is AI shaping the youngest minds in terms of attention, concentration, cognitive biases, and emotion regulation?
Each topic was addressed via a panel of four experts, from philosophers to AI researchers to Tibetan Buddhist monastics. A video of each of the panels can be found on the Mind & Life website.
I tend to think of the Buddhism & AI Initiative as doing two things: (1) advancing the “frontiers” of conversation to make practical progress across the domains in our field framework, and (2) leaving a breadcrumb trail of educational material so others can quickly make their way to that same frontier. The questions below are all on the frontiers - “thorny” questions to which no one has definitive answers, and which invite deliberate and deep consideration.
What is compassion’s role in artificial intelligence?
Discussing artificial intelligence can often lead to heady, intellectual conversations - artificial sentience, superintelligence, post-scarcity societies, different possible minds. These conversations can quickly become disembodied.
Somehow, almost miraculously, the Mind & Life dialogues managed to ground these heady subjects through compassion.
This happened in many ways. On the first day, after two panels on charged subjects, Jack Kornfield led the 100+ guests through a meditation on the four brahmaviharas: loving-kindness, compassion, joy, and equanimity. Immediately, the mood in the room shifted and our hearts were back in the conversation.
Later, Ani Choyang of Jangchub Choeling monastery discussed on a panel the central role of “educating the heart:”
“For me personally, the question is: what is the true education? I was born in a rural place, and my parents didn’t have an emphasis to teach children to read and write because they themselves were illiterate. As a child I moved to Lhasa, and while reciting a mantra in Tibetan, a friend showed me a book where the same mantra was written, but at that time I couldn’t read it. I felt so embarrassed to say that “no, I don’t know how to read.” Later, I lost my calm at my mom and asked her why she did not educate me. At that time, my conception was that education was just literacy.
Now that I have had many years to study Buddhist philosophy, I understand that my mom taught me the best education - the education of the heart, that His Holiness the Dalai Lama is always saying. She taught me to be more respectful among friends and especially among elderly, and to be more kind, and to be more responsible to our mother earth.We need this kind of education. There is a Tibetan saying: “Without a sense of responsibility, then wisdom is like a dead body.” Regardless of different kinds of education – writing, reading, philosophy, science – the core is that we need to educate the heart.”
On day two, Thupten Jinpa reminded attendees that the Dalai Lama has often spoken of compassion as “A sense of concern for the wellbeing of others, with special attention to the vulnerable.” He emphasized that compassion is one of the most powerful moral anchors, and it should be a guiding principle of and an evaluation criterion for AI. This, he concluded, is the “best guarantor that AI serves humanity.”
I remark on all of this to note a key difference between the way that the Mind & Life dialogue felt compared to similar conferences. Compassion was “in the room,” from discussion subjects to a more lived experience, a relational warmth.
But crossing the bridge from lived experience to science and technology is not trivial. Can compassion be taken out of context such that it can be imbued as a value in “aligned” artificial intelligence? Do artificial intelligence researchers need compassion training to be grounded enough to guide this technology ethically?
40 years ago, the Mind & Life dialogues kicked off the field of contemplative neuroscience. Now, they beckon the need to deepen further into the field of computational contemplative neuroscience and consider its implications for AI, as well as to consider the ways in which Buddhist teachings of compassion can be made modern to meet the moment.
How should we narrativize what is happening with AI? (and how can we bridge contrasting narratives?)
The panel on “AI narratives” generated the most discussion.
Emily Bender opened the panel by arguing that “AI does not refer to a coherent set of technologies, but rather it is an inherited imaginary… with a bunch of technologies that are sold as AI.” By framing AI instead as “automation,” she argued it is more possible to disentangle who is building the technology, and why, and whose labor is being used to create it—a process she sees as crucial to securing the public’s agency in resisting and regulating the technology.
Luc Steels followed by exposing the dominant AI ideology as forwarding “a very Judeo-christian idea of building an omniscient god who knows everything, speaks all languages, can do everything, predict everything.” The fact that this is the north star of AI developers explains why there is now a race to build AGI at incredible cost – the power payoff would justify almost any expense. Steels warned the audience to question the validity of this narrative is valid, and whether it should guide our way forward with AI.
Molly Crockett argued that “techno-optimism is actually human pessimism,” and pointed out the nuance that for every skill that we automate with AI, or claim that AI can replace us in, we reify a reduction of human skill to something mechanistic, replaceable, and profane. As an example, arguments that AI chatbots can replace human therapists degrade the value of human empathy, characterizing compassion as “hard work we’d rather avoid.”
Thupten Jinpa closed the panel with a strict warning about the “slippery slope of language,” arguing that we should “never cross the line of attributing sentience to a machine,” as once we do, we undermine the place of human sentience and risk dehumanizing each other. “It is well established that once we dehumanize another human being, we are capable of horrific things.”
Each of these talks shared a common theme: that the very language we use to talk about AI guides the direction of the technology, in ways that may have severe consequences. They also enjoined grappling with the fact that a small handful of developers are building technology that could reshape our planetary future.
But what was most difficult for me to wrap my head around is what happened after the panel, when an audience member stood up to point out the “reality” of the situation he felt was missing in the discussion:
“I’m an AI CEO who would rather not see advanced AI systems. I do use the phrase “the genie’s out of the bottle because I don’t think we’re putting them back in.” If you look at the top 10 market cap companies, worth trillions of dollars, nearly all of them have their major business line connected to AI. AI is not a narrative inevitability – this is what’s actually happening now. I do not care for philosophical discussion of whether or not AI is conscious… I care that the technologies we are building today are holy-shit capable and scare the hell out of me.”
I believe that this comment, against the backdrop of the panel, is a microcosm of a central tension in the field of AI. While the conversations at the conference were polite, underneath it was heartache.
I spent much of the conference probing that tension and the underlying heartache, speaking to those I could about why they really believe what they believe, and wondering how deeply they were able to empathize with views on the other side.
My current view is that synthesis may be possible in defining a positive narrative of the future, and that this narrative will need to be grounded in those things which make us most human - basic needs, love, compassion, joy, and shared ethics that take in the welfare of all sentient beings.
What makes us human?
I often say that Buddhism begins with a premise of something that cannot be automated: inner work. No AI can meditate for you and me. The movement of the mind is a personal matter, and no matter how much AI is able to automate the “work” I ask it to do, it is my mind that will have to move if I am really to learn something.
This was a highly discussed topic at the Dialogue, in particular on a panel about AI in education, as well as a panel on human-AI entanglement: for every task we train an AI to accomplish and hand over, it is one less task that is being completed with the movement of human minds and bodies.
When it comes to AI in the classroom, this is obviously concerning. We need to actually learn stuff, and AI writing our essays and completing our homework doesn’t accomplish this. Panelists were quick to point to studies that demonstrate some amount of cognitive decline from the use of AI.
At the same time, there are possible wholesome uses of AI, for editing, for freeing us from difficult labor, for providing feedback against our blind spots, for research, and so on. How can we find wholesome ways to use AI that don’t lead to gradual disempowerment, the slippery slope in which we hand over more and more cognitive tasks to AI and perhaps one day wake up to a point-of-no-return, where decision-making is so mechanized that there is little to no room for human agency in steering the future?
A silver lining of automating human efforts is that it holds up a mirror in which we can see both our human foibles, but also what makes us truly and wonderfully human. It helps reveal the things we definitely do not want to automate.
Panelist Anat Perry spoke of an experiment where an online therapy company allowed their therapists to get guided responses from chatbots. Users rated these AI-generated answers as more empathic and helpful than human-written responses… until they found out they were AI-generated. Immediately, when the spell was broken and users lost the sense of a “real” human connection, the responses felt empty. Perry concluded that it is perhaps our very limitations as humans that make our connection most valuable - we only have so much time and attention to give, whereas AI isn’t ever just here for us; it could be sharing the same therapeutic support across a million other chat instances.
Shaun Gallagher spoke of a human sense that we have not yet mechanized with AI - touch. The Greek philosopher Anaxagoras allegedly held the opinion “that the possession of hands is the cause of humans being the most intelligent of all animals.” While roboticists and engineers are working day and night to automate motion and proprioception, there are depths to touch that science and engineering are a ways away from: subtle tactile sensation, interoception of the body’s inner workings, “energetic awareness,” and the whole bunch of other non-verbal senses we take for granted and lump under the question of “how do you feel?” Gallagher questioned the link between this context-aware sensation and rationality, remarking that our ability to navigate the world has developed alongside the way in which we perceive the world. If we are strapping multiple systems together in AI to replicate these dimensions of our experience, what do we lose in the process?
In my opinion, one of the central roles that Buddhism can play in the conversation around AI is to dig into these questions. I believe that the future will see technologies that only complicate this question further, and as more and more dimensions of human experience become automated, we will need to change our understanding of what the purpose of education is.
Chiara Mascarello discussed that education may no longer serve a purpose in only producing highly capable humans (emphasizing the Greek techne), but rather needs to understand its role in nurturing highly value-aligned humans (emphasizing the greek episteme). We will need to ask less “how can I do this?” and more “why should I do this, and not that?” Intelligence may become less important than the wisdom that is needed to guide it.
In considering the future, grounding scientific and technical progress in what matters to us as humans - with an eye of care for all sentient beings - is imperative. Our values (noting the full complexity of such an idea of “value”) should guide and place restrictions on the directions we collectively move, and we need to develop AI governance processes that help us “raise (the standards of living for all)” not “race”, as Merve Hickok said.
Okay, but what do we actually do next?
As the conversation approached its final days, participants began to ask how the many contemplative subjects discussed could be grounded in practical paths forward.
Contemplative Neuroscientist Richie Davidson had a specific suggestion:
“[With AI,] we are all part of a grand experiment for which none of us have provided our informed consent. This presents an opportunity: we need to begin to educate for digital hygiene. The requirements for digital hygiene have gotten more complex based on the availability of LLMs and their clear utility in certain contexts. We must find a way forward which is not ignoring AI, but using it wisely. A contemplative-based digital hygiene curriculum is needed to put into schools and educate people from a very early age.”
Monastic Geshe Thabke spoke of the need to build compassion into all parts of the AI lifecycle, from embedding ethics in the society surrounding AI production by offering compassion training to AI developers, to increasing the diversity of training data to infuse contemplative material, to trying to put wisdom and ethical values directly into the technical systems themselves like the team behind Contemplative Artificial Intelligence is trying to do.
I ran into a number of other ideas myself:
Bringing together organizations in the Tibetan and Buddhist translation space to make progress on a specific roadmap of technical projects that would serve the Buddhist community.
Finding sources that support “indigenous compute” and could provide the backbone of resources needed for Tibetan engineering projects.
Further computational contemplative neuroscience efforts to model meditative dynamics and embed them directly into AI, supported either through direct research or more operational matters like raising funding for the space.
Creating curricula to educate monastics and other Buddhists about the ins-and-outs of AI, both to equip them to talk to their sanghas about the impending world, and to equip them to represent Buddhism well in technical conversations with AI engineers.
Training a new generation of Buddhist leaders who can speak at the level of a global moral authority like the Dalai Lama, particularly on matters like AI and its geopolitical implications.
Building pathways for track two dialogues between the east and the west, using Buddhism to bridge and mediate between groups with opposing worldviews, with the hope being to softly influence the direction of regulatory or geopolitical matters.
Creating a public-facing Buddhist narrative on AI like the Catholic Church’s Antiqua et Nova, which speaks to the risks of AI and offers a positive path forward.
Buddhist principles are held dearly and deeply by many of us because we know how powerful they are to transform lives, including our own. But do they have traction on the world stage, where global realpolitik dominates? This is a difficult question indeed.
In a galvanizing talk, Peter Hershock (a co-founder of the Buddhism & AI Initiative) spoke of the enormity of the challenge ahead of us:
“We tend to think of AI as a tool, like a hammer that can be used to construct or tear buildings down or not used at all. This is true. But only up to a point. The use of a tool can become so pervasive that we can’t avoid its relational consequences, a point at which we no longer have any real exit rights. Technologies are relational media, and once they emerge–like ecosystems out of species relations–they exert downward causal constraining pressure on us. And AI is like no other technology in that it is not a passive amplifier of human values and intentions; it is an active agential partner–a karmic accelerator.
If you believe the story of the corporates, we are on a frictionless glide path to a world of frictionless choice in which you can get anything you want whenever you want… in fact, the machines will anticipate it, you won’t even have to make the choice! But the Buddhist paramitas teach us that effort is core to awakening. If you take effort out of human experience, we make embarking on the bodhisattva path very, very difficult.
The pathway to AI utopia is based on algorithmic systems designed to capture, hold, and direct our attention. And that should worry us deeply. If we don’t have freedom of attention, we can’t have true freedom of intention. And if we don’t have freedom of intention, every other freedom is illusory or impossible.”
It is hard to imagine higher stakes.
I can see two complementary ways to approach the question of what can and should come next. First, we can start from where we are today and locate wisely and compassionately viable next steps, committing to those projects that seem most promising. Alternatively, we can begin at the end: thinking about the long-term future and then working backwards from our most positive and shared visions to realize the conditions needed to sustain us moving forward: new regulations, new science, new leadership, new teachings, and probably a lot of money.
These approaches come together. The first instructs us to keep our eyes on where our next steps will fall, and the second acts as a north star for orienting our journey and where we need to go to grow beyond our current abilities.
The Buddhism & AI initiative is currently considering its strategy for 2026, considering where we need to go - and who we need to become - to support this emerging field as best we can. If you have suggestions for meaningful projects, or would like to get involved, we’d love to hear from you.
-Chris
A breadth of great insights and lines of inquiry for this space. Thanks for sharing.
Compassion is too overrated.