Zen and the Art of Intelligent Robotics
Ben Goertzel Interviews Chinese Zen Buddhist AI Researcher Zhou Changle
In the last few decades, science has unveiled a tremendous amount about how the mind works – exposing various aspects of the hidden logic of cognition, perception and action, and connecting the mind to the brain. But modern science is not the first discipline to probe the nature of the mind. Multiple traditions within diverse cultures throughout history have probed the mind in their own ways. For example, my teenage self spent many perplexed hours poring over books on Buddhist psychology, reviewing the theories of the 14th century Indian logicians Dharmakirti and Dignaga, whose detailed analysis of the different possible states of consciousness seemed in many ways more revealing than modern consciousness science.
Among the numerous approaches to the mind, Zen Buddhism ranks as one of the most intriguing. Far less theoretical than modern science or medieval Indian logic, Zen emphasizes experiential wisdom and the attainment of special states of consciousness – Zen “enlightenment.” While a great deal has been written about Zen, this is a literature fraught with irony, since one of the central notions of Zen is that true understanding of the mind and world can’t be transmitted in language.
According to the traditional story, Zen began one day when Buddha, addressing an audience gathered to hear his teachings, chose to remain silent instead. Instead of discoursing as expected, he stood there quietly holding a flower in his hand. Most of the audience was confused, but one disciple, Kashyapa, experienced a flash of sudden understanding – and gave the Buddha a subtle smile. Buddha gave Kashyapa the flower. Then Buddha finally spoke to the crowd, “All that can be given in words I have given to you; but with this flower, I give Kashyapa the key to all of the teachings.” This is the origin of Zen — a “wordless transmission” of mind to mind. And – to continue the traditional tale — 1300 years or so later, in 500AD, the Indian monk Bodhidharma brought the wordless teaching to China, founding Chan, the Chinese version of Zen … which later spawned the somewhat different Japanese version of Zen, that is perhaps the best known species of Zen in the West.
Zen is famous for brief riddles called koans, such as “What is the sound of one hand clapping?” Meditating on such riddles has the potential to bring enlightenment, when the mind suddenly realizes the futility of approaching the koan (or the world or the mind) with conceptual thought, and fully embraces the paradox at the heart of the koan (and the world and the mind).
Chinese Zen spawned what has been my favorite statement about Zen since I first encountered it when I was 18 years old, in the wonderful little book The Zen Teachings of Huang Po
The dharma of the dharma is that there are no dharmas, yet that this dharma of no-dharma is in itself a dharma; and now that the no-dharma dharma has been transmitted, how can the dharma of the dharma be a dharma?
The word “dharma” means, roughly, the essential teachings of the Buddha (which lead to enlightenment) and the essential constituent factors of the world.
Huang Po gives us Zen straight up, not prettied-up for Western consumption like some of the Western Zen literature (e.g. Robert Pirsig’s famous and excellent book Zen in the Art of Motorcycle Maintenance which inspired the title of this interview).
And Zen is alive and well in China today, as I found out for myself recently. For the last few years I’ve been collaborating on AI and robotics research with a lab at Xiamen University in south China (see my earlier article on the Chinese Singularity for a little more about this); and on one of my visits to Xiamen, I was delighted to learn that Zhou Changle, the dean of Cognitive Science at Xiamen University, is a very accomplished Zen practitioner.
Changle was one of the original masterminds of the Artificial Brain project that brought my good friend Hugo de Garis to Xiamen a few years ago. Hugo retired recently, but the project continues within Changle’s BLISS (Brain-Like Intelligent SystemS) lab. We have 3 humanoid Nao robots in the BLISS lab, together with some supercomputing equipment and a host of capable grad students and professors, and are working on making the robots intelligent and (in at least some sense) self-aware. Our technical approach involves a combination of the OpenCog AI architecture I’ve co-created, Hugo de Garis’s evolving neural net approach to perception processing, and the ideas of Changle and his Chinese collaborators.
Hugo’s main interest, in the context of our Xiamen project, regarded large-scale brain simulation using neural nets on specialized hardware. On the other hand, Changle’s scientific contribution to the project has more to do with machine consciousness and the logic of mental self-reflection – topics that, after I learned of his involvement in Zen, seemed to fit in perfectly. And beyond our collaborative AI/robotics project, Changle also carries out research on a host of other topics including computational brain modeling, computational modeling of analogy and metaphor and creativity, computational musicology and information processing of data regarding traditional Chinese medicine.
Changle originally pursued his study of Zen under a teacher – but after several years of that, his teacher judged he had reached a sufficient level of enlightenment that he should continue on his own! And as well as practicing Zen and various forms of science, he has written extensively (in Chinese) about the relation between the two, including over a dozen papers and a book titled TO DEMONSTRATE ZEN— ZEN THOUGHT VIEWED FROM SCIENCE.
The abstract of one of his recent talks gives a flavor of how he connects Zen with cognitive science:
The various states of consciousness experienced by human beings may be divided into four categories, using the philosophical concept of intentionality: intentional, non-intentional, meta-intentional, and de-intentional states. Analyzing Zen “enlightenment” in the light of this categorization, one concludes that Zen thinking is a de-intentional self-reflective mental capacity. This establishes a philosophical basis for the Zen method of mind training, enabling the exploration of connections between Zen, philosophy of mind, cognitive science, and other areas.
— however, Changle’s talks are much more colorful than this conveys, filled with humorous stories in the rich Zen tradition.
I interviewed Changle recently about Zen and robotics and consciousness, focusing specifically on some areas where we don’t quite agree, regarding the relationship between AI and consciousness.
Changle foresees AIs and robots may eventually become extraordinarily capable, potentially even launching some sort of technological Singularity. But he also thinks they will always be impaired compared to humans in their internal experience, due to a lack of a certain type of “consciousness” that humans possess – a “consciousness” closely related to Zen enlightenment.
On the other hand, I myself suspect that AIs and robots can eventually become just as intelligent and conscious as we humans are. And ultimately, I suspect, they can become just as enlightened as any human — or more so.
Unsurprisingly, during our conversation we failed to come to full agreement on this set of issues – but we did have an interesting dialogue!
Changle, I’ve interviewed a number of scientists and also some spiritually-oriented people — but you’re the first person I’ve interviewed who’s BOTH an extremely accomplished scientist AND deeply experienced as a spiritual practitioner within a particular wisdom tradition (Zen). I’m curious how you balance these two aspects of your mind and life. Do you ever find them pulling you in two different directions?
Zen, strictly speaking, is not a religion in the sense that Buddhism, Christianity and Islam are. It is a philosophy towards life. Thus it won’t collide with science. In the purest sense of Zen, it doesn’t even emphasize meditation. Freedom of mind and enlightened wisdom are what matter.
It’s an intriguingly paradoxical philosophy towards life – and its elusiveness seems part of its essence. I remember you gave a talk about Zen and science entitled “Not Two.” And later I saw the Chinese characters for “Not Two” on an archway in a big Buddhist temple complex on Hainan island.
Zen basically goes against any forms of dualism. A is true; A is false; A is true and false; A is not true and not false.
So once you’ve described Zen in some way, you’ve set up a dualism between the things that match your description, and the things that don’t. And so what you’re describing isn’t really Zen, which is non-dual…. It’s simple to the intuition yet perplexing to the thinking mind. I get it. And I don’t get it. And I both get it and don’t get it!
Heh…. But I remember once you said that only Chinese could become enlightened! So I guess I don’t get it.
That was a joke, of course!
Of course — but the reason it was a funny joke is that there’s a little bit of truth at the core of it.
Yes, the intuitive Chinese mentality and the holistic Chinese philosophy have advantages over the Western analytic mentality of dualism, in terms of grasping the essence of Zen. But even so, Zen is still very elusive to most Chinese.
Now let’s turn to your AI research, and its relationship to your Zen intuitions.
Superficially, there would seem to be a certain disharmony between the views of the mind implicit in AI research and Zen. AI is about taking parts of the material world, and arranging them in such a way that intelligent behavior results. But Zen takes a very different view of the mind – it views the mind as more like the basic ground of existence, not as something that comes out of arrangements of material objects like semiconductors or switches or neurons. How do you reconcile these two positions?
To me, the material world reveals itself through the human mind. In other words, it never exists without a human understanding it. However, mind is not an entity, but rather a process of making sense of everything. The intelligence that we are achieving through robots is not an entity either, but rather a process of demonstrating intelligence via the robot.
But what about consciousness? Let’s say we build an intelligent computer program that is as smart as a human, or smarter. Could it possibly be conscious? Would it necessarily be conscious? Would its degree and/or type of consciousness depend on its internal structures and dynamics, as well as its
Consciousness, as I understand it, has three properties: self-referential, coherent and qualia. Even if a robot becomes a zombie that acts just like it possesses consciousness, it won’t really possess consciousness — because the three properties of consciousness cannot all be realized by the reductive, analytic approach, whereas AI is based on this sort of approach.
Hmmm… but I wonder why you think the three properties of consciousness can’t be demonstrated in a robot? I agree that human-like consciousness has three properties (self-reference, coherence and qualia) – there are many other ways to describe the properties of consciousness, but that seems OK to me. But as a working assumption, I’d be strongly tempted to assume robot can have all of these….
I wonder which of the three properties of consciousness you think a robot is incapable of having? Qualia, coherence or self-reference? And why is it that you think a robot can’t have them?
To me, a human can be thought of as a kind of “meat robot” — so I don’t see why a robot can’t have all the properties a human can have.
Though I have to admit my own view of consciousness is rather eccentric. Although I don’t identify as a Buddhist, I share with many Buddhists the notion of panpsychism – the notion that “everything is conscious” in a sense.
But of course I can see the “reflective, deliberative consciousness” that humans have is something different than the universal consciousness that’s immanent in everything including rocks and particles. I wonder what’s your view on the relation between these two types of consciousness — on the one hand, the “raw consciousness” that we panpsychists see everywhere; and on the other hand, the “reflective, deliberative consciousness” that humans have a lot of, mice may have a little of, and rocks seem to have basically none of.
In Zen we say that SUCHNESS is the nature of everything. For a conscious human being, its Suchness is its consciousness; but for material things, their Suchness is their properties such as shape, substance or weight but not consciousness.
Hmm, the word “suchness” seems hard to translate from Chinese to English!!
Turning to the Internet for guidance, a writer named A.H. Almaas says that:
In simple terms, to experience ourselves as Being is to experience our existence as such, to experience our own Presence, our own “suchness” directly. It is the simplest, most obvious, most taken-for-granted perception that we exist. But this existence is usually inferred, mediated through mind, as in Descartes’s “Cogito ergo sum” – “I think, therefore I am.” Existence is background, not foreground, for our ordinary experience. To penetrate into this background, to question our assumptions about reality and ourselves, allows us to encounter directly the immense mystery of the arising of our consciousness and of the world.
It is Suchness, pure Suchness. We cannot say anything about it. We cannot say it is self, we cannot say it is not self, we cannot say it is God, we cannot say it is the universe, we cannot say it is a person, we cannot say it is not a person; the moment we say anything, we are within mind. If we use any concept here, even the concept of purity, simplicity, or whatever else, we are within the mind, and we are blocking that which cannot be named.
The Chinese word for “suchness” is “真如” or “自性”, i.e. “buddhahood”, also known as “the foundational 识” or “种子识” or “Alayavijnana” etc…. It is the name for the nature of everything.
So you say everything has Suchness, but only for certain systems like humans does the Suchness manifest itself as consciousness per se. I’m reminded of some Western philosophers like Stuart Hameroff, who say that everything has “proto-consciousness”, but only in some specific cases like people or other advanced animals does the proto-consciousness manifest itself as consciousness.
Whereas, I would say that everything has “consciousness” but only in some specific cases like people or other advanced animals or sufficiently sophisticated AIs, does this basic consciousness manifest itself as “reflective, deliberative consciousness.”
But it’s actually quite hard to tell the extent to which we disagree on this point, versus just having very different terminologies. Human language was created to discuss practical matters, not to have precise discussions about subtle points like this… and translation between English and Chinese makes it even trickier.
I’ll explain my view in a little more detail.
According to the Buddhist perspective, mind is a kind of process, not a kind of entity. The process of mind contains: five kandhas (also called aggregates) and eight senses (八识) which are also often thought of as “eight consciousness.” The eight senses are:
- Eye (vision/visual sense)
- Nose (olfactory sense)
- Ear (auditory sense)
- Tongue (taste sense)
- Body (tactile sense)
- Mind (consciousness)
- Manas (realization/awareness)
- Alaya (seed, prajna, suchness)
Note: the last one is the “suchness”!
If you want to compare these Buddhist concepts with “consciousness”, the sixth sense (“真如”, translated as Mind above) is the most similar with the western concept of “consciousness”. But according to Zen, Manas (the seventh sense, i.e. the ability of awareness or realization) is more nature than consciousness; and the only thing which can decide the Manas is Alaya (Suchness). So really, the concept of “真如” doesn’t exist in the system of Western philosophy – it’s not exactly the Western concept of “mind” or “consciousness.”
And corresponding to these eight senses, the five aggregates identified in Buddhism are:
- Matter or Form (rupa): Perception, i.e. the physical form corresponding to the first five organs of the eight senses, which has intentionality. AI robots can have this ability
- Sensation or Feeling (vedana): The feeling in reception of physical things by the senses through the mind or body; emotion, also known as “qualia”, which doesn’t have intentionality. It’s hard to give AI robots this kind of ability, but probably not impossible.
- Thinking (senjna): Thought and cognition, which has intentionality.AI robots can have this ability
- Volition or Mental Formation (samskara): Conditioned responses to objects of experience, such as language, behavior, ideas and soon, possessing intentionality. Robots can have this ability.
- 识: Awareness/Realization/Subconsciousness, which is anti-intentionality-ed / de-intentionality-ed / un-intentionality-ed. And this is beyond the scope of robots!
Regarding the project of implementing the “mental power” of humans in robots, the key troublesome point is number 5 in the list — the implementation of “awareness mechanics” (i.e. the ability of de-intentionality), which is the so-called enlightenment of Zen. For this purpose, current AI research, which is based on a reductive and analytical approach, is not helpful at all.
I think I understand what you’re saying; and I find this perspective fascinating. And I know the details run a lot deeper than what you’ve described just now; you’ve kept it compact and simple since this is an interview not a long lecture….
But still – as you probably expected, I’m not convinced.
Once again … why do you say that a robot cannot have 识 (awareness, de-intentionality,…)?
I mean, think about people for a moment. From one perspective, humans are meat — we’re made of organs, cells, molecules, etc. From another perspective, humans are minds that have 识 …
Similarly, from one perspective, robots are machines — they’re made of wires, memory disks, silicon molecules…. But from another perspective, perhaps robots can eventually be minds that have 识, just like humans…
After all, Ben cannot directly experience Changle’s 识 … just like Ben cannot directly experience a robot’s 识. But even so, Changle’s 识 has its own suchness, and perhaps the robot’s 识 can have its own suchness as well — if the robot is made correctly! Say if it’s built using OpenCog….
The awareness (enlightenment) of Zen is beyond all concepts, but all the approaches we use for building robots and AI systems, and all the behaviors of robots and AI systems, are based on concepts
Heh…. I had intended to ask you sometime during this interview whether you think an AI program or an intelligent robot could ever experience enlightenment in the same sense that a human can. Can a robot attain enlightenment ?– can we have a Zen Master bot? But I think your remarks have made your view on this pretty clear.
Right. Robots can’t be enlightened because they don’t possess consciousness.
Your view has an impressive coherency to it. But even so, I’m going to keep on being a stubbornly unenlightened Westerner! My perspective is more like this:
- We may initially construct an AI robot according to our plans and ideas and knowledge – our concepts, as you say
- However, once we release the robot into the world, and it starts interacting with the environment and self-organizing, then the robot becomes its own being … it goes beyond our initial programming, just as an embryo grows beyond the original cells that make it up, and just as a baby eventually goes beyond its initial neural wiring. Through its own growth, self-organization, development and interaction with the world, the robot’s “suchness” can develop consciousness, insight, human-like qualia and knowledge over time.
At least, that is my current belief and hypothesis. But we will need to build MUCH better AI systems and robots to find out if I am correct or not!
And of course, that’s what we’re working on together in the BLISS lab in Xiamen…
Anyway it’s good to hear your different perspective, articulated so well. Life would get boring if everybody agreed with me…
I don’t think we’re going to get much further on the question of consciousness just now, so let me shift attention to a slightly different question, still in the robotics vein. I wonder just how sensitively you think humanlike intelligence and consciousness depend on having a humanlike body.
For instance, in our lab in Xiamen we have 3 Nao robots, which our students are controlling with AI algorithms. The Nao robots are roughly humanoid looking, but they have many shortcomings relative to human bodies — for instance, their skin has no sensation in it; and they can’t feel any sensations inside their body; they can’t move their faces in accordance with their emotions; etc. Setting aside subtle issues of consciousness — do you think these physical shortcomings of the Nao robots will intrinsically prevent an AI embodied by the Nao robot from having a human-level intelligence? How realistic do you think a robot body would have to be to really support human-like intelligence?
The human intelligence and body are not independent of each other. If they are viewed separately, that is dualism. The robot’s AI and the robot’s body are not independent of each other either.
OK – so one last thing, then. You’re familiar with Ray Kurzweil and Vernor Vinge’s notion of the Singularity — a point in time at which technological change becomes so rapid that it’s basically infinite from the human perspective, and superhuman AIs rather than humans are the main driving force on the planet. Kurzweil foresees it will come about in 2045. What do you think? Will a Singularity happen this century? Will a Singularity happen at all? If a Singularity does happen, what will it mean in Zen terms? Could a Singularity possibly mean Enlightenment for everyone, via for example special computer chips that we plug into our brains and that stimulate the brain’s spiritual centers in a way that spiritually “wakes a person up”?
Such technology breakthroughs are possible. But putting computer chips in the brain might not produce novel conscious experiences, if consciousness is discovered to be based on quantum entanglement.
Hmmm…. Once again you remind me of Stuart Hameroff a bit – what he calls proto-consciousness is closely related to Zen “suchness”, and he’s quite convinced that human consciousness depends on macroscopic quantum phenomena in the brain.
However, again, I’m a bit unconvinced. I think there are some subtle points about quantum entanglement that no one really understands yet. For instance, there’s some mathematics suggesting that quantum entanglement is really about the relationship between a system and the observer, not just about the system in itself. So, in some cases a system typically modeled using classical physics, may be best modeled (with respect to a given observer) using quantum mathematics and quantum entanglement. I wrote a blog post about this a while back; maybe you can have a look when you get the chance.
Anyway, Zen enlightenment is an experience of a certain attitude, which a robot or brain-implant computer chip will have a hard time to experience as it doesn’t possess consciousness. So it’s not entirely clear how AI or brain chips can help with enlightenment, though they can do a lot of interesting things.
Well, if human consciousness is related to quantum entanglement, then maybe we could enhance human consciousness and ease human enlightenment using quantum computer brain chip implants! But I guess we’re going pretty far out here, and I’ve taken up a lot of your time – it’s probably a good point to draw the interview to a close.
One thing that comes through clearly from your comments is that the Buddhist perspective on mind is a very rich and deep one, which nudges AI and robotics research in a certain direction, different from how Western philosophy habitually nudges it. For instance, Western AI theorists are constantly arguing about the importance of embodiment for intelligence, but for you it follows immediately from nonduality. And the Buddhist theory of mind gives you a very clear perspective on which aspects of human intelligence can be expected to be achievable in AI robots. I don’t agree with all your views but I think AI and robotics can use a lot of different ideas and perspectives. So I hope we can get more and more cooperation between Western and Chinese AI and robotics research, and get more and more cross-pollination and synergy between the different approaches to mind.
ACKNOWLEDGEMENT: This interview was conducted in a complex mish-mash of English and Chinese. Copious thanks are due to Xiaojun Ding and Ruiting Lian for assistance with translation, including translation of some rather subtly untranslatable Buddhist philosophical terms (some with roots in Sanskrit as well as Chinese).