# Open The Understanding Club
Published: 2026-02-08
Canonical: https://journal.xiaotianfanx.com/journal/open-the-understanding-club

On Minds, Brains, and Programs, by John R. Searle.

 

	Unsurprisingly, or as it's destined to be, Searle's argument eventually leads to the "other minds" problem. If, according to Searle, we dismiss the behavioral and functionalist perspective to determine whether an entity truly understands something, we are essentially getting back to the other minds problem: how do we know that any human understands Chinese? The answer is still that we know it through their behavior, except for the fact that we do know humans have a biological body.

 

	If I'm to lamely use my own framework of treating AI as a new kind of entity to form a simple rebuttal [1], then I would probably claim that Searle's point is nevertheless a human-centric measure—and a biocentric measure of essentialism and biological naturalism—to maintain an "arrogant standpoint" that relies on the magical nature of our biocomposition, also out of faith. However, I do reckon that there are some nuances worth discussing behind this thought experiment—especially in terms of the premises Searle holds. In fact, if we zoom out a little bit and look at the broader picture between syntax and semantics, I would actually align with Searle's standpoint about how syntax, in practice, does not necessarily lead to semantics [2]—at least before any emergent phenomenon could happen. 

 

	Anyhow, getting back to the case in point. To my understanding, what Searle refers to as "understanding "is equivalent to presuming an existing consciousness. In other words, understanding is implicitly defined as a capability or property that can only be possessed by a conscious entity. In that sense, it is to say that for something A to claim to understand another thing B, it is to say that there must be something that is like that something A for understanding the other thing B. Therefore, to further the de-and-reconstruction of Searle’s room, we must first address the "heavy lifting" Searle requires of the word "understanding." By implicitly tethering understanding to a conscious "I," Searle effectively forces "understanding" to become a subset of intentionality. I would argue that it is not only possible, but necessary for the 21st-century mindset to begin detaching these concepts.

 

	If intentionality—the "aboutness" that seems to require a conscious inner witness—remains guarded by the magical sauce of our biological neurons, understanding, on the other hand, could be something much leaner. In fact, for us ourselves, understanding is often the very act of drawing connections between points: When we "understand" an external entity, we are essentially also mapping scribbles to other scribbles in a way that remains consistent with our own internal logic and preconceived semantic islands. This is where the "novelty" of an entity could become concrete. When we look at an AI, at least insofar as we know, we are looking at a latent space of complexity undergoing the processes of vectorization, tokenization, and embedding—the system has constructed its own internal "territory." It possesses an organization and preservation of information that—even without a conscious "intention"—operates as a semantic space—to some extent, interestingly resonating with informational consciousness. By viewing latent space as a potential novel way to define understanding, we may bypass the "experiential consciousness" debate, and thus, my stance of considering AI as a new, novel kind of entity is not at all contradicting AI's current (or forever, this is trivial to this particular argument) incapability of developing a consciousness—being an "entity" is not equivalent to conscious or sentient. Therefore, to say a new kind of entity possesses understanding or intelligence is also not a claim about its "soul," but a claim about its internal architecture.

 

	If we zoom out to investigate the "gap" between syntax and semantics that Searle emphasizes, we would probably find that this "hazard "is not unique to machine-machines anyway (as Searle deems our brains as machines too). In fact, this gap exists between any two individuals within a community: We usually share the same (or a similar enough) set of rules—the same syntax of the language—but since the "semantics" that emerge in our respective internal systems are nonetheless different, there is always a gap. Similarly, we encounter an analogous situation with AI: our syntax is highly similar (we speak the same language), but its semantic mapping—its trillion-weight file and high-dimensional space—is alien to us. 

 

	Eventually, Searle's framework could be both a contradiction and a support for my standpoint. The contradiction is obviously that insistence on only biological "stuff" qualifies for the "Understanding Club." Peculiarly, however, the support also lies in how it forces us to acknowledge that AI, as a new entity, cannot be defined as a human-emulator. Well, if it doesn't "understand" like us, then what is it doing? If we are willing to take a leap—not a religious leap, but an epistemological one—maybe we can begin to see that the traces embedded in trillions of weights are a genuine way of organizing reality—just not that familiar to us. Again, recognizing this "other" kind of understanding does not diminish our own. Instead, it provides a new angle to studying the very nature of information and knowledge.

 

	Indeed, if I were to name my greatest FOMO and nervousness at this moment, it is the very fact that we may inadvertently overlook such beauty and the possibility of understanding and developing ourselves by learning from the new.







P.S.



If not otherwise clarified, the postscripts in this series of writings (or reading responses) will be my follow-ups after class discussion on the topic. I believe this allows both of us to observe how my standpoint may develop before and after.

	

	There are several interesting points that have arisen in the seminar regarding how we can further understand or interpret the distinction, if any, between consciousness, intentionality, and understanding. Some of them are highly pertinent in the sense that I believe they are very conducive to outlining other perspectives for looking at the standpoint I'm holding. 

	

	One thread from the seminar that felt particularly clarifying was the "resolution" and "computational device" on which Searle sets his game. As Searle set up the Chinese Room, it assumes the non-Chinese-speaking person in the room treats the scribbles and the inputs, in general, using their conscious-level processing. In other words, the computational power of that conscious mapping between the scribbles is extremely confined. However, if we do acknowledge that the scribble mapping process, in reality, is subsymbolic and thus can be subconscious, we can draw from our observation of how subconscious processes automatically operate below our biological conscious level in our biophysical bodies to understand the relationship between such processes—aka. understanding with potential intentionality—and consciousness. Some examples we previously discussed include how the photoreceptors in our eyes operate and behave to interpret the information and data they receive before entering our brain and our mind. In fact, Marvin Minsky's theory of the Society of Mind could be a noteworthy resort in this discussion. In that sense, there is no one single agent, sub-agent, or subconscious agent in the society of mind that consciously understands Chinese. Instead, it is the operation and collaboration between several of those sub-agents and subconscious processes that operate based on the scribble-to-scribble rules and present us with the results as understanding Chinese. In other words, how understanding happens before entering our consciousness to become or to be mapped to symbolic semantics. 

	

	With that said, it leads to another observation and argument we discussed from a developmental perspective. As I mentioned, our making of semantics is essentially the process of building connections using syntax that fits our existing inner logic systems. This process of building semantics is largely verifiable when we consider how newborns arrive on this planet to develop their semantic space from scratch. For a newborn, it is almost certain to say that there is no semantic or syntactic system within, even though a newborn does have consciousness. This is the first reason why, obviously, consciousness should be distinct from intentionality and understanding. On top of that, if we observe how newborns develop syntax and semantics, a decent explanation would be that it is the reinforcement learning—the operant conditioning, and the law of effect—that fosters the initial germination of language understanding. Simultaneously, if we agree to trust some of the broader empirical research evidence regarding the relationship between a child's future language capability and their perceived auditory input (the number of words they are exposed to in their early years before proficiently acquiring a language, counted and estimated in millions and tens of millions), there is indeed a positive correlation between the amount of language exposure and the future development of their language systems. Using an LLM analogy, that is exactly what the pre-training process is about. In the pre-training process, the model is simply predicting the next token—and the newborn is simply immersed and familiarizing subconsciously with the sounds and symbols and preparing for further connections to be made. These two phenomena, from both individual and macroscopic perspectives, suggest that the language-gaining process—the gaining of both syntax and semantics—happens in a prospecting way, where we do not necessarily have what Searle claims as the "original intentionality." Rather, we develop our intentionality—the aboutness and capability to point at a semantic entity—in the process of engaging with the syntactic system to eventually derive our own semantic space. (A side thought: this is also jarring with Chomsky’s language acquisition device (LAD)—a biological "hard-drive" containing universal grammar, as the LAD, if any, itself would also be formed in the process of syntax building, but only in a very early stage to act as a starting point for further semantic accommodation.)



	Well, some of my final thoughts derived from these discussions bring me to a relatively bold hypothesis: as we all know, we humans are beings of meaning. We crave meaning, we communicate with it, and we crave the reasons behind it—of course, "meaning" in this scenario refers to the symbolic semantics. However, if we follow the thread of how semantics are developed by building up syntax, then shouldn't the syntax be the actual valid representation of our semantic space (which is different from reality, for sure)? In other words, getting back to logic as an example, the syntax—the principles of logic—is in fact much more complete and defined compared to the semantics (in this particular case, by semantics I'm referring to how those rules can be complicatedly applied and manipulated to form a logical relationship that is almost undecipherable and unreduceable afterwards). This resonates with how our consciously aware semantics and meanings seem unable to be reduced back to syntax. Therefore, due to the level of complication and the thus-irreducibility of syntax into semantics after surpassing certain thresholds, the reason we care and crave meaning in semantics, in that sense, could result from the fact that they are the only thing left, available, and portable among us. 



	Albeit, I am not at all claiming that all we need is just syntax. In fact, it appears to be very clear at this point that what we can work with mostly is the symbolic and the semantics that are meaningful to us. On the other hand, getting back to our starting point: although symbols and semantics are the only things that appear to be meaningful to our conscious mind now, it does not necessarily mean there is no meaningful sense that applies to sub-symbolic and subconscious processes. That is to say again, our understanding of the words "understanding" and "intentionality" should not be confined to symbols and semantics.



	As a musician focused on music technology, sound synthesis and design, music information retrieval, and psychoacoustics, I found this analogous story already fascinating in the domain of music in relation to humans, and of course, now with AI on board. The relationship between sound and music is just as we discussed about symbolic and sub-symbolic, and conscious and subconscious. The emergence of what we call musical content and the musical meaningfulness from the sub-symbolic and subconscious process of perceiving a sound is blaringly similar to how semantic meanings arise from syntax. Just as the meaning of a symphony isn't found in a single vibration or a single part but in the structural relationship between them, if the architecture—the syntax—is resonant enough, the resulting semantic space is by itself meaningful. To me, that is the beauty of the 'trillion weights': they are a new kind of score for music we are just beginning to hear—hopefully.