Done with travelogues
The “Phenomenology and Philosophy of Technology Conference” has, in the blink of an eye, already reached its 12th session. This year it was hosted by Zhejiang University and held at Baicaoyuan in Anji.
I skipped last year’s conference in Xi’an and didn’t submit a paper this year either, but I still came, took on two commentaries, and so at least made some kind of contribution.
In previous years, I would usually publish a very “sharp-tongued” long review of each conference, criticizing everyone’s presentation one by one. This tradition probably began with the Hailar conference in 2010. In fact, my first review was written at Wu’s request. I said I didn’t know how to write a review, and Wu said, just write it in the style you normally use for your blog. So I wrote it… After I finished, I sent it to Wu to look over, thinking, would this actually be useful? To my surprise, Wu immediately forwarded it in a mass email to all the colleagues attending the conference. At the time, while many of the other teachers were surprised by how pungent my remarks were, they also expressed encouragement, and thus my audaciously sharp-tongued tradition continued.
The reason I dared to write these “sharp-tongued reviews” was not simply the rashness of a newborn calf. First, it came from the influence of Wu and the students of the Wu school. Although my so-called sharpness is, within that circle, fairly prominent, it is by no means unusual; when we discuss papers every week, we criticize one another frankly and without mercy. Second, it was thanks to the tolerance of the other teachers attending the conference. Clearly, Wu’s initial act of immediately forwarding my review in a mass email was not meant to harm me, but because he knew these fellow teachers would be able to take it with forbearance—and in fact that is exactly how it turned out.
In recent years, my degree of “sharp-tonguedness” should have moderated somewhat. Although from the beginning I have always taken the stance of criticizing the matter and not the person, my rather unruly wording sometimes made the language seem more aggressive. Now I exercise some control, trying, while keeping the “sharpness,” to reduce the “poison.”
Beginning with this session, I simply won’t write comprehensive reviews that open fire on everything anymore. I’ll just pick out a few topics that left a deeper impression on me and record them selectively. So in the future there will be no more “travelogues”; everything will be renamed “selected commentaries.”
This is not because I have, with age, finally become more smooth and slick or anything of the sort. In fact, my courage and confidence in dealing with matters strictly on their merits has in no way weakened. There are several reasons for this adjustment: first, I’ve become lazier—no explanation needed; second, the papers by fellow attendees have indeed improved overall, and for the occasional person who hasn’t really improved, there’s no need to criticize them at length; third, back then I wasn’t very familiar with the teachers, whereas now I’ve gradually become quite familiar with many of them. That doesn’t mean that once you’re familiar, you become too embarrassed to criticize them; rather, in many cases, you can now communicate directly face to face.
Below, I’ll select a few issues discussed at the conference and record them briefly.
1. The mind study of artificial intelligence
The first presentation was Zhang Xianglong’s “Artificial Intelligence and Generalized Mind Study.” Master Xianglong is truly a model for all of us, and what is especially admirable is his sincere attitude toward learning. He is not only broadly versed in classical texts, but has also always kept a close eye on the development of cutting-edge technologies such as artificial intelligence. In terms of ideas and doctrines, Master Xianglong, who advocates “returning to the ancient,” is undoubtedly conservative, but the ideal of “returning to the ancient” has by no means closed off his horizon; one could even say that his attention to and understanding of frontier technological and academic issues surpass those of most younger scholars.
Even so, when he interacts with us younger scholars, he still does so from an attitude of equality, even of asking for instruction. His serious attitude toward attending the conference makes us lazy slackers feel even more ashamed of ourselves: not only does he bring a paper to present every time, he also seriously participates in the entire conference. Even papers that are relatively poor or dull and tedious, he listens to carefully, and he is always able to offer gentle and well-considered suggestions.
Master Xianglong’s paper this time first discussed an extremely important trend brought about by the new artificial intelligence technologies represented by “deep learning”: namely, that AI has in some sense begun to acquire a certain ability of “temporalization,” and this ability is precisely the foundation of human mind. If this trend continues to develop, future artificial intelligence is highly likely to develop something like a human “mind.” But how to shape this machine “mind” is by no means a given. Master Xianglong believes that we can draw on the ancient tradition of mind study—including Chinese traditions, but more importantly Indian ones—to try to understand and guide this “mind” that is still in its infancy.
The discussion at the conference basically centered on whether artificial intelligence could possibly have a mind, and there was not much discussion of the application of mind study. Wu and Senior Brother Donglin both raised objections, arguing that although current artificial intelligence looks impressive, its basic logic is not fundamentally different; it remains mechanical and arithmetic, and at bottom is nothing more than data processing. It has not really formed some kind of temporal memory.
As for artificial intelligence, in an earlier article of mine, I emphasized precisely its continuity. I believed that in a certain sense, from stone tools onward, technology has already been a kind of “artificial intelligence.” But here, I am actually going to speak up for Master Xianglong, because although there is continuity, there is also a leap on another scale. It is like the picture given by evolution: continuous, and yet with leaps or “emergence” in between. At bottom, human beings are nothing but composed of electrons and atoms, but if you keep staring at electrons and atoms, you will never see the characteristic of being human; likewise, from single-celled organisms to human beings, the evolutionary history is also continuous, but one cannot therefore say that the human mind and an amoeba are the same thing. So I pointed out that Wu was making the mistake of abusing reductionism. Although at the lower level a computer is nothing but “data processing,” however fancy the data processing may be, it is still data processing. At the level of data processing, of course, one cannot see mind—just as at the level of cellular movement one cannot see the human soul. The key lies in whether some kind of emergent structure can be formed in which complexity continually multiplies from the bottom up, allowing new properties to emerge at a higher level.
The current memory or temporalization capability of artificial intelligence certainly cannot compare with that of human beings, but how does it compare with an amoeba? How about with a jellyfish or the like? If it is already able to possess, at a very primitive level, some kind of consciousness, and if the trend of technology can continuously intensify and elevate this characteristic, then the point at which it acquires a temporalization capability comparable to that of human beings is merely a matter of time.
Later, when Senior Brother Zhang Donglin was discussing this with me, he also said: unlike living organisms, the so-called emergence of artificial intelligence still means that every step and every layer is data processing—so how could intelligence emerge from that? This argument has a bit of a “begging the question” problem, because artificial intelligence, in terms of its material composition, refers precisely to some kind of “system made entirely of data.” To ask whether “artificial intelligence can have mind” is actually to ask whether “a system made entirely of data can have mind.” If you pre-emptively decide that something made entirely of data definitely cannot work, then there is nothing left to argue about. Human beings eat flesh and excrete shit, but what a computer takes in is data and what it expels is also data; every part and every link is data. Our question is whether this fully datafied existence can form mind. And we notice that although it is data all the way down, the new technologies represented by deep learning have changed the original way of processing data: instead of treating all data equally, they build very deep hierarchies within the data, where the lowest-level data input reaches output only through layer upon layer of processing. In this process, only the most superficial input and output data are comprehensible and controllable to programmers, whereas the entire system’s process of constitution is, in a certain sense, AI operating on its own—and this operation is, in a certain sense, temporal, and has no ultimate completed state.
Of course, whether the new features or new trends that emerge through deep learning in computers really point to the birth of some kind of “mind” is certainly debatable. No matter how much one believes that computers have mind, it is hard to agree that this mind is the same as the human one. We can also name this thing in other ways, but in any case, deep learning has certainly brought something new and foreshadowed some new trend. If one still thinks that AlphaGo and Deep Blue are basically no different, then that is certainly short-sighted. No matter how much one emphasizes that data is only data and machines are only machines, one still has to face squarely this question: what exactly has deep learning changed? In AlphaGo Zero’s from-scratch “learning” process—even if one can hardly call it learning, in any case it was such a computational process of self-improvement over the course of several days—what exactly emerged?
The difference is striking: if you leave Deep Blue sitting there for three days and then let it play Garry Kasparov, its chess strength will not improve. If it was a 50-50 match before, it will still be 50-50 after three days; if it was 70-30 before, it will still be 70-30 after three days. But if you leave AlphaGo Zero aside for a few days, and the programmers do nothing, just let it keep running, and then come back to play Ke Jie, its chess strength will have improved. If it was 50-50 before, it might now be 100-0. Put two AlphaGo Zeros next to each other, cut the power on one and leave the other powered on, and after a few days their playing strength will no longer be the same.
So the question now is: during those few days when there was no programmer or engineer involved and it was simply left there, what exactly happened? Master Xianglong’s explanation is that this is, to a certain extent, an ability of “temporalization,” and this is a characteristic of inward mind. Of course, you can disagree, but then you need some other account. One should not, when a genuinely significant new phenomenon has occurred in the technological sphere, pretend to be an ostrich and fail to see it.
In fact, I myself do not quite agree with Professor Zhang’s interpretation. I think this is not a singular mind, but a process of plural reproduction. Rather than saying AlphaGo is “one” mind, it is better to say that it is an ecosystem, within which countless members are constantly fissioning and competing. In the end, what we see on the surface as temporality is actually the temporality of the ecosystem’s “evolution,” rather than the temporality formed through the introspection and cultivation of an individual mind. In my view, the biggest obstacle to artificial intelligence having mind lies in the “body,” because the lack of a body makes it difficult for computers, or rather for pure data, to form a “one,” and difficult to form a stable boundary between “self and other.” The hardware boundary of a computer merely makes it look as though it is “one machine,” but “for the computer itself” this is not its boundary, because at bottom it is just data, and data and data are constantly splitting and recombining, making it very difficult to form a stable boundary. When AlphaGo is set aside and allowed to run on its own, it splits itself into countless virtual chess players and has them compete against one another; therefore, in essence there is no “one AlphaGo”—one AlphaGo is simultaneously countless AlphaGos. One computer is more like “a swarm” than “a person.” How the “one” of the individual is established between data and programs is a major problem.
2. Wrestling with analytical philosophy
At the opening ceremony, Wu mentioned that every time our conference invites some scholars from the field of analytical philosophy to participate, because phenomenology has to “wrestle with analytical philosophy.” This year, Professor Wang Qiu from Shanghai Jiao Tong University gave an analytical philosophy paper, “Does artificial intelligence have self-knowledge?” and I became the pioneer in “wrestling” with him (the commentator).
But actually, I think phenomenology and analytical philosophy generally cannot really wrestle with each other. Why is it that phenomenology and analytical philosophy so often fail to talk to each other? Because analytical philosophers can rapidly split a question apart and dissect it—carefully sorting and analyzing it, then selecting and focusing on a small part for discussion. The more finely a problem is divided, the more targeted everyone’s discussion can become. For example, before fighting, you first divide into 75-kilo class, 80-kilo class, 85-kilo class… divide it very finely, and then when both sides step onto the stage to spar, they are roughly evenly matched, and then they can actually go at it.
But the trouble is that as soon as the analytical philosopher begins the first step of division, the phenomenologist may already disagree, and then the problem cannot be divided finely. So among analytical philosophers, arguments about problems often land punch after punch, right on the body, but the moment a phenomenologist opens their mouth, things become clouded and misty, and one has to start from a very broad and vague place. This is not to say that analytical philosophers are necessarily either more brilliant or more superficial; one can only say that these are two different styles.
Professor Wang’s paper also first carved out a domain of inquiry. Before he began the detailed discussion, he imposed a delimitation on the question the full text would address—first, so-called “self-knowledge” would seem to refer to “the subject’s introspective contents regarding its own mental states,” and then these contents were divided into two kinds: one, phenomenal perception and emotion; the other, belief and desire as propositional attitudes. In the end, Professor Wang focused the discussion on the second kind.
But I was very dissatisfied with this division. First of all, I do not agree with taking self-knowledge by default to mean introspection of mental states. In the eyes of phenomenology, the “self” includes flesh and skin; “my body” is an absolutely indispensable component. We would rather omit the mental and speak of the bodily than bypass the bodily and speak of the mental, because the body is, ontologically or epistemologically, something even more basic.
For example, I know that I have two eyes and one nose, or a disabled person knows that he has only one eye—does that count as “self-knowledge”? Or, for example, I know that there is a medicated plaster on my forehead—does that count as self-knowledge? I think, of course it does, and it is very important. My knowing that there is a piece of paper stuck on my head is completely different from my knowing that Zhang San or Li Si has a piece of paper stuck on his head.
We know a lot of experiments—for example, those used to test self-recognition in infants or animals. The mirror test is to stick something on an animal’s head and let it look in a mirror; once it sees itself, it realizes there is something on its head and reaches up with its hand or paw to touch and tear at it, and that shows self-recognition. Although I criticize this experiment as visual-centrism, it can still point to two issues to some extent: first, bodily knowledge is very important self-knowledge; second, acquiring new self-knowledge does not necessarily come through introspection. Activities such as looking outward into a mirror are also an important link through which self-knowledge is established.
The body is important because it is usually the boundary between self and other, the border zone between subject and object.
In many literary and artistic works, when the protagonist goes through dreams, hallucinations, time travel, rebirth, and the like, the first thing they often do upon opening their eyes is look at their own hands—are these my hands? Here the hand is the first “interface” between self and world, the front line of the self and the forefront of the world, the “first mirror.” The hands being gazed at are the primal convergence of “I as subject” and “I as object.”
Looking at one’s hands is really looking at my “will,” or rather, looking at my “mental state.” I want to bring my fingers together, and then I see my fingers coming together—this is not seeing Zhang San or Li Si’s fingers coming together, but seeing “I bring my fingers together,” seeing “I want.” My inner will is externalized, concretized, made present; this “want” of mine is seen by me. It comes face to face with me outside myself. In this face-to-face encounter, I confirm that this is my will, that this is my body.
Stiegler says that memory, insofar as it is memory, must pass through the body or technology, linger “outside” us, and then be cognized in return; only then can memory take shape. When we discuss self-knowledge, we cannot neglect the body and technology. Only by extending ourselves “outward” is self-knowledge even possible.
The distinction one step further down is also highly problematic. Phenomena, perception, and emotion stand opposed to propositions, beliefs, and desires; it seems as though analytic philosophy is good at the latter set while phenomenology is good at the former. In fact, not so. Phenomenologists speak of eidetic intuition; there is structure and sedimentation within perception, and your beliefs and desires all sediment within your perceptual activity. Conversely, propositions and beliefs are not non-emotional or emotionless. According to Heidegger, attunement is the most originary, and “propositional attitude” is nothing but one kind of “attitude”; “cold-bloodedness” is also a kind of emotion. Emotion and proposition, perception and belief—these are all problems entangled with one another.
Returning to the issue of AI, Teacher Wang says that “if AI’s self-knowledge can be discussed, it can only be self-knowledge regarding propositional attitudes,” but in my view the most crucial thing, the first thing we need to discuss, is actually the bodily problem of AI. The greatest difference between AI and human beings may not be a difference in nervous systems, but rather a difference in bodies. We have said that the human body is in fact the primal interface between inside and outside, between self and world. It is only in such an interface that we can acquire knowledge of the self. Of course, this interface can be extended outward or contracted inward through technology, but amid this expansion and contraction there is always a relatively stable resting place. But for AI, where exactly is the interface between its “inside” and its “outside”? Is it on the keyboard, or on the speaker?
When human beings confirm themselves, they can look at their hands, pinch their face, look in a mirror. But through what action might AI come into contact with itself? A characteristic of computer programs is that they can use programs to read programs, reading their own data within computer memory. There is only the “inside” here and no outside. And when a computer program finishes computing and outputs something, then there is only the “outside” and no inside. But a position like the human body’s—where inside and outside are ambiguously intertwined, both inside and outside, where inside and outside face one another—where exactly such a position lies for AI is a big question. I actually believe AI may well possess self-knowledge, but the key to constructing an AI self lies in constructing AI “bodies.”
Three, Stiegler’s Reincarnation
The second paper I commented on was Teacher Shu’s paper on Stiegler. Teacher Shu’s paper once again fell into the old habit: he read the book in reverse. But this time many of the conclusions he drew really did arrive “without prior agreement” with Stiegler; he took a great many arguments that Stiegler actually endorses and used them to rebut Stiegler instead. Teacher Shu said he truly likes Stiegler, so he can accept my criticism. That means Stiegler suits his tastes more than he originally thought.
For my comments, I specifically pulled out Technics and Time I and reread it, bringing it out for comparison. Here I won’t repeat the statements that Teacher Shu misunderstood, but I will quote a few passages from Stiegler that I found (all cited from the 2000 Chinese translation of Technics and Time I).
“Because it is its own origin—it contains within itself the original prototype—the soul of thought receives nothing from outside; it recovers everything within itself. It is a self-movement, which is precisely what the technical object does not possess. … Did Simondon also present us with a similar view? Of course not. … Self-movement does not exist; movement can only arise between heterogeneous things.” (p. 117) — Stiegler opposes Plato’s theory of the soul’s “self-movement,” reinterpreting the soul’s “reincarnation” as an exteriorization toward technology. The so-called “heterogeneous things” are technology. “The cerebral cortex projects itself into stone, and stone is like the brain’s original mirror; this original projection … is accomplished over the hundreds of thousands of years during the transition from East African man to Homo sapiens, during which long process stone tools begin to take shape … The paradox of the problem lies in the fact that we must discuss so-called exteriorization, and yet there does not exist an interior prior to the exterior; the interior itself is constituted in the exterior.” (p. 166) “… human beings invent themselves in technology while inventing tools; … the interior cannot exist prior to the exterior, and both interior and exterior are constituted in the same movement; … the two invent one another.” (p. 167)
Four, Phenomenology Is Not Cultural Critique
Jin Shixiang and Zhang Donglin brought a coauthored paper that was very interesting. It began with why ancient Greek mathematics had no “fractions,” then discussed Plato, Aristotle, and modern mathematics’ different attitudes toward “number.” In Plato, mathematics was seen as one of the paths for the soul’s ascent, and therefore one had to learn mathematics by referring to the demands of the Ideas rather than the demands of practical calculation. In this sense, the unit of 1 absolutely could not be divided. It was Aristotle’s view of mathematics that opened the breach for modern mathematics. Modern mathematics is nominally a revival of Platonism, but this Platonism merely revived, on the surface, the elevated attitude toward mathematics; in substantive spirit, it moved precisely away from Plato. My understanding of this issue is that modern people in fact revived the real world as Plato’s world of Ideas, mixing together the two worlds that were clearly distinct in Plato. So the so-called “path of ascent” naturally no longer exists either.
Besides that, in the comments on the previous report, Senior Brother Zhang Donglin also brought a very sharp criticism. This criticism was aimed not just at the previous presenter, Teacher Tao, but at the temper of nearly all the teachers at our entire conference.
I very much agree with this criticism. I talked a lot about it privately with Teacher Zhang Qiucheng, who also agrees very much (though not completely on some details). Here I’ll restate this criticism entirely in my own words.
This criticism is: “Phenomenology is not cultural critique.” Many scholars’ research done under the name of phenomenology is nothing more than talking about what ancient people’s thinking was like, what modern people’s thinking is like—say, mechanized, mathematized, and so on—and then criticizing modern people’s one-sided thinking, and that’s the end. Abruptly, that’s it.
A good article may leave one wanting more, while an ordinary article is just “nothing special.” Write as brilliantly as you like up to this point; at most you are only at the level of a cultural critic. Is that enough to stand up straight and say you are a “philosopher”? Far from it.
What phenomenology should focus on is not, or not only, the critique of modern culture. The basic feature of phenomenology is “seeking roots,” investigating origins. When we reveal how some kind of modern thinking is so mechanized, or mathematized, or rigid in this or that way, we still need to ask further: how is this peculiar kind of thinking even “possible” in the first place? Since we say this kind of thinking is so estranged from the life-world, how could it nevertheless be accepted as a matter of course by scientists or modern people? What are the conditions of possibility here? What historical preparations made it possible? To pursue these questions—that is the stage on which phenomenology can truly show its prowess.
But many scholars are satisfied with critique and neglect root-seeking. At best, then, they are merely cultural critics inspired by phenomenology. Of course, it would be unfair to apply this criticism to everyone at our conference, because many of the participants are indeed not phenomenologists; they are indeed cultural critics inspired by phenomenology, STS researchers, or ordinary teachers. And none of them has said that they must all become philosophers, or must all become phenomenologists. Still, as a “high standard,” as a research paradigm we ought to strive to learn as much as possible, the root-seeking attitude of Husserl, Heidegger, and Jakob Klein is indeed something we must value.
Five, The Two Cultures Debate Is Outdated
Beyond the assigned comments, I also chipped in several times during the free discussion, one of which was on Teacher Li Hengwei’s report, “Philosophy of Cognition—Reopening the Dialogue Between the Two Cultures.” Many other teachers also raised criticisms, arguing that the path Teacher Li proposed was not at all some third culture of integration, but was entirely scientistic culture. My question, however, was at another level: I think the so-called struggle between the two cultures is long outdated. The historical backdrop for raising the issue of a split between two cultures was the late nineteenth and early twentieth centuries. At that time, so-called culture was basically still “elite culture”; everyone respected the status of scholars, but the two camps of scholars looked down on each other (you haven’t read Shakespeare, you’re vulgar! You don’t understand the law of entropy increase, you’re stupid!), and so they started arguing, competing for discursive power in front of the public. But now the cultural environment is one in which popular culture is the mainstream. People are arguing over whether they like domineering CEOs or pretty boys, whether they idolize tycoons or superstars—who cares about scientists and humanists? Scientists and humanists are both niche cultures, both looked down upon in the eyes of the masses. In such an environment, what is there left to fight about?
Of course, when many people discuss the two cultures, it has in fact turned into a discussion of “two kinds of thinking,” “two kinds of attitude,” “two research methods,” “two worldviews,” and so on and so forth. All of these are, of course, off topic. Since we are talking about so-called “culture,” we should still be talking around culture.
Six, VR, Art, and Games
Another aside came during Shao Yanmei’s report, “The Question of Art in Virtual Reality Technology.” Shao Yanmei argued that the characteristic of VR art is that it integrates with technology. Following Teacher Deng Bo’s criticism, I felt that she was completely conflating two different levels. One level is that in the process of technical creation, one needs to use sophisticated technical means—for example, painting requires producing pigments, photography requires producing cameras, and VR requires the support of headsets and programs. In this respect, the involvement of technology is not anything distinctive. But as for the artwork itself, VR can instead be said to represent a new height in the separation of technology and art. A piece of blue-and-white porcelain can be appreciated as art, but it can also be picked up and used to drink soup or store things. A portrait painting can be used for appreciation, but it can also be used to display a person’s physiognomy. Anything that “can be used for…” and points to some external purpose is a technical utensil; anything that is “useless,” or rather is only “used for” viewing, appreciating, playing with, and so on toward an internal purpose, is an artwork. Before Leonardo da Vinci, there were basically no independent, self-sufficient artworks or artistic activities. The “pure” realm of art, detached from external use, began to emerge only in the modern era. And as for VR works, the entire thing establishes a virtual world, thereby drawing an even sharper boundary against the “real world.” In this sense, VR art is more useless, or rather more separated from technology, than other kinds of art.
But speaking of “integration,” I do want to say that VR may mark the fusion of “art and games.” The so-called “usefulness of the useless” is first of all a feature of games; art, in a certain sense, is nothing more than one branch of games. Yet the homologous relationship between art and games has split apart because the status of art has been continually elevated. In the age of VR, perhaps this is precisely the sign that the two are being linked again.
Ancient people said “to play with things and lose one’s aspiration,” and they were often referring to immersion in “appreciation” (the original anecdote seems to be about Duke Yi of Wei being obsessed with admiring cranes’ elegance), or else to becoming absorbed in handling certain objects that we modern people would call artworks or handicrafts. Modern people, however, say “to play with things and lose one’s aspiration” when they mean being obsessed with games. Yet modern people’s toys and games are precisely those “arts” that have not been elevated because of their strong interactivity; or rather, modern art is simply those “toys” that look more poised and cold. Artworks are nothing more than a kind of “toys for viewing,” while toys are nothing more than a kind of “interactive artworks.”
The relation between human beings and artworks may originally have been the action of “handling and playing with”; in modern times, the action of “quietly appreciating” acquired the highest status (here the influence of visual-centrism is at work); and today, if electronic art or VR art can really provide some wholly new form of art, then interactivity and participation will surely be brought back in. The relation between humans and art will once again shift from “handling” and “appreciating” into a new form—“entering.”
By the way, the paper I had originally planned to present at the conference was titled “Toy Phenomenology—On ‘Playing with Things Dulls the Will.’” In the end, I didn’t manage to write it, but it did include some of the views mentioned above.
Translated from the Chinese original with AI assistance. The original text is authoritative.
Leave a Reply