Well, as it turned out, although the model learned the orbits with high empirical accuracy, when they looked into the actual math equations that were derived, they did not resemble Newton’s laws at all. They were a total garbled and inconsistent mess, resembling something like epicycles, the pre-Copernican theory of planetary motion based on circles orbiting circles orbiting circles, rather than the elegance of Newton. So what gives?

In fact, this result should not be surprising at all; it should have been obvious as soon as the research questions were raised. Neural networks and epicycles are not only analogous, they’re actually different implementations of the same idea: To approximate something, you can do so by building an iterated system where at each step you correct the errors of the previous steps via a simple mathematical mechanism. Neural networks implement this idea via a sequence of layers, each of which consists of simple elements called perceptrons. Each layer corrects the errors of the previous assemblage of layers. The “deep” of “deep learning” refers to the use of a large number of such layers. Epicycles implement this idea via a sequence of circles, each of which corrects the errors of the previous assemblage of circles.

Given such a system, the a priori assumption should not be that it yields simple solutions, but that it yields complex ones. This is almost definitionally true, as the word complex literally means composed of many parts. Neural networks are composed of many layers, and epicycles are composed of many cycles. This is why the result that the neural network learns “laws” more resembling epicycles than Newton should be completely unsurprising.

Expecting a deep neural network to derive Newton’s inverse-square law is like expecting an epicycle model to stumble onto heliocentrism.

Now, there is nothing intrinsically wrong about the use of approximation systems; complexity on its own isn’t necessarily better or worse. Such systems are used to great effect all over mathematics and engineering. A canonical example is Fourier series — sums of sines and cosines — which form the backbone of our entire telecommunications infrastructure. Indeed, epicycles can be thought of as an early form of Fourier analysis.

The only problem arises when we think of these approximations as providing metaphysical laws, rather than giving us what exactly we asked for, which is, again, just empirically accurate approximation.

We may denigrate epicycles as pseudoscience today, but in fact they were, and remain, a surprisingly accurate model of planetary motion. In Structure of Scientific Revolutions, Thomas Kuhn argued that the main reason to discard epicycles and geocentrism was not for reasons of empirical accuracy, but rather for human reasons — because Newtonian physics and heliocentrism is a much simpler conceptual theory requiring only a single notion of gravitational force, rather than a complicated notion of orbits around orbits around orbits, requiring correction after correction after correction.

That is to say, the empirical accuracy of a theory may or may not correspond to other desirable properties of the theory, such correctness, simplicity, generalizability to new contexts, or human understandability.

The story with LLMs mirrors much of the above discussion although with some subtle differences.

First of all, natural language is different from other domains because it is a universal substrate. Planetary orbit data is just about planetary orbits. But natural language data includes essentially everything. It contains data about planetary orbits, too, in addition to data about every other conceivable topic, such as nuclear power plants, dinosaurs, causal inference, Avril Lavigne, and the history of Ancient Greece.

Neural networks and epicycles try to infer an underlying structure behind the data they are given. When the data is just planetary orbits, they will learn a set of “laws” for the orbits, which is likely to be more complex than necessary, as explained above. But when the data is all the text on the internet, then the “laws” that they derive must be more robust, because these “laws” need to conform with a larger set of phenomena, not just isolated phenomena like planetary motion — or so the argument goes.

Indeed, this argument does seem to hold in many instances! A simple example is grammar. Have you ever seen an LLM make a grammatical error? It happens — but it’s very rare. Grammar is a “latent variable” behind most of the text that the LLM sees during training. This means that grammar influences the text, but it is not explicitly given to the model. Same as the influence of Newton’s laws on planetary motion. The LLM must derive grammar as an underlying “law”. The fact that grammatical errors made by LLMs are so rare, even in novel scenarios, seems to indicate that the LLMs have in fact derived the correct grammatical laws.

But the argument is decidedly not true in other instances, especially for factual information. Hallucinations are epicyclic. For example, when you ask for citations and you get a list of research papers that don’t exist, it’s because the model has imagined a false universe in which those papers do exist. Research papers are factual entities which do not have any particular structure or law that underlies their existence. When they exist, they exist by a combination of random chance and happenstance, as a matter of what happens to be funded and what happens to be discovered by researchers. So in this case, the model tries to find structure where there is none. It derives a theory of epicycles for the existence of research papers, when a better theory would be no theory at all.

Where does this leave us in our assessment of LLMs and AI more broadly? Are they truly intelligent, understanding the true metaphysical nature of reality, or are they just epicyclic snake oil?

Sorry to be boring, but there is no straightforward general answer. The model sometimes learns a correct underlying world model, and other times it makes up a nonsensical epicyclic model. There is as yet no real way to control this. There is as yet no real way to predict the correctness of the world model in any given domain.

The main thing we can say about a deep neural network is that it should achieve high empirical accuracy in any domain that is well-represented in its training data. Whether they can go beyond that statement, whether they can usefully generalize, can’t be known in advance. The only way to find out is again empirical, to just try things and observe the results.

In many regards, I find LLM therapy (Claude specifically; ChatGPT sucks for therapy) better than human therapy. However, it’s important to use it in the “right” way. And the right way to use it is not in a “chat” fashion, with short back-and-forth messages. Instead, you should write long stream-of-consciousness braindumps, providing as much context as possible. Basically, you should just treat it like a journal that happens to talk back to you.

One reason to send long messages is because if you send short messages, the model generally will be biased to send short messages back and remain in a “chatty” mode. In particular, it will avoid doing deeper analysis. But if you send a big long message with a lot of context, it will tend to do more holistic analysis and be more “thoughtful”.

From the technical side of things, it seems likely to me that RLHF tuning primarily focuses on single-turn interactions – after all, single-turn data is easier to acquire and easier to evaluate than multi-turn data. As such, I think that the models likely perform better in fewer-turn interactions. So that’s why I think long messages work better than breaking things into many short messages sent over many turns. But this is just a hypothesis; I work in AI but not in LLMs specifically so I have no direct knowledge.

Another strategy for Claude therapy is to use Claude’s “Projects” feature. Generally try to keep each session focused on one specific topic. When a session feels like it’s gotten too long or it feels like it’s reached a conclusion, ask Claude to write a detailed summary. Then copy that summary into a new document which you upload to that Project, so that future chats have it in the context. As you keep doing this, the benefits compound, as the model now has more and more context about you, and can make deeper and deeper connections in its analysis. This is essentially the same as ChatGPT’s “memory”, except that you have control over what is included in the memory.

You have to be somewhat careful, though, and you can’t just blindly trust that the summaries are correct. They are sometimes inaccurate, which will taint your future sessions. But that’s also true of a human therapist, whose memory will be imperfect. In both cases, you’ll have to correct the therapist. For Claude, you may also have to go back and manually edit the context document. You still have to put in work; there is still no free lunch.

As for the therapy itself: One of the functions of a therapist (but not the only function) is just to be a sounding board, an interlocutor, who is able to understand what you’re saying without judgment, and reflect it back to you, perhaps in a cleaner or more organized form. For many people, including myself, the act of verbally discussing my thoughts through many iterations is the best way to process something. When you have a more practically-oriented problem, LLMs work just as well as a human therapist, if not better, because they can suggest essentially every possible common strategy for your problem. So working through practical problems with the LLM therapist to find potential solutions is a much more efficient process.

On the other hand, an LLM therapist may not work as well on emotionally-oriented problems, where you may need deeper empathy from a human (or the illusion of such empathy). Another important part of human therapy is transference: Transferring feelings you have about other people onto the therapist. The relationship with the therapist, who is also a human with all the same human flaws that we all have, becomes a sort of virtual stand-in for your other relationships, and a place where those human dynamics can be played out.

The LLM is not human. It is superhuman: it’s infinitely patient, it’s super fast, and it knows about almost everything. It’s possible to make the LLM pretend to not be superhuman, but it’d nevertheless not be its “natural” state. So it seems unlikely or at least very difficult to play out human relationship dynamics – anger, jealousy, resentment, disappointment, love, tenderness, affection, inspiration, mentorship, and so on – with an LLM.

Nevertheless, having an 24/7 on-demand superhuman therapist is an amazing tool that cannot be undersold. At worst, LLM therapy is like having instantaneous access to every self-help book ever written. No self-help book on its own will ever directly solve any of your problems, but the tools they contain are still valuable and often not something you can come up with independently on your own. The LLM can help you to become aware of such tools and personalize them to your unique context. You still have to make use of them.

• Why is your favorite movie your favorite movie?

• I don’t remember the first time I saw Lost in Translation. I’ve probably watched it two dozen times or more.

• Half of the times were on airplanes, where Lost in Translation was for some reason often available in the airplane movie library for a good decade or so. I guess it’s because the movie has a travel theme, and you travel on airplanes. I almost always watch Lost in Translation when it’s available on a flight.

• It hasn’t been available on any of my flights over the past couple years; now I generally rewatch The Matrix or Challengers instead.

• I’m in my late 30s, and I feel a bit jaded about travel now. But I was an enthusiastic traveler in my 20s. I romanticized the concept of travel and the serendipitous adventures and romantic flings that would — or at least could — happen while in the liminal space that is travel. That’s one reason why I liked Lost in Translation so much. That’s also why I liked Before Sunrise.

• I stopped liking Before Sunrise in my 30s; I felt like I’d outgrown it. I began to see Ethan Hawke’s character as just a pretentious, corny, pseudo-intellectual young guy, probably the same way I was in my 20s. Cringe.

• I’ve continued to enjoy Lost in Translation into my 30s. One reason was that I found the party scenes to be very realistic portrayals of fun, drunken nights out with friends, getting into trouble, running around from one place to the next, having brief conversations with a bunch of random characters, going crazy at karaoke, falling asleep on the taxi ride back with city lights in the background. I love having nights out like that. So these vignettes evoke strong nostalgic feelings for me.

• I have to note that Lost in Translation is kind of racist. It does bother me. It’s just ignorant. You can’t just brush this under the rug as “the characters are racist, it doesn’t mean the film is racist”. The racism is a choice by the director. The film spends an inordinate amount of time fixated on accents and linguistic misunderstandings for no good reason. Haha, Japanese people can’t speak Engrish, haha. Ok, fuck you.

• One of my favorite factoids about Lost in Translation is that it’s meant to be a fictionalized version of Sofia Coppola’s breakup with Spike Jonze, where Charlotte (Scarlett Johansson) is Coppola and Charlotte’s husband (Giovanni Ribisi) is Jonze; presumably Bob (Bill Murray) is fictional. What’s funny is that Jonze made a movie, Her, which is also a fictionalized version of his breakup with Coppola, and which also features Johansson, this time as a fictional manic pixie AI gf. Which makes Bill Murray the manic pixie AI bf of Lost in Translation.

• Another reason why I like the movie because I can relate to the characters’ ennui and feelings of “what the hell am I doing with my life”. Bob and Charlotte both feel this way, despite their difference in age. I felt this way when I was Charlotte’s age, and I expect I’ll still feel this way when I’m Bob’s age, too. These are existential feelings that are not fixed by being in a relationship or career success.

• The soundtrack is superb. Back when I lived in California, I used to listen to it all the time while driving. The soundtrack introduced me to so much good music. It introduced me to Kevin Shields, which led me to My Bloody Valentine and shoegaze. It introduced me to Haruomi Hosono and Happy End, which led me to Yellow Magic Orchestra and Ryuichi Sakamoto. Sofia Coppola has great taste in music.

• I became a father in June 2024. It’s changed everything. It’s even changed how I view Lost in Translation.

• I rewatched Lost in Translation in December 2024. I was surprised that my interpretation of the film dramatically changed in that viewing, and I saw things that I never saw before. Here’s what I wrote:

I rewatched Lost in Translation again last night for the 1000th time. I used to think of it as the “Bill Murray as manic pixie dream man” movie; now I still think of it that way, but with some extra layers of complexity.

Bob is unhappy in his marriage, but he is able to experience youthful fun again with Charlotte while in Japan. Ultimately however, it’s an escapist, impossible fantasy that can only briefly exist in a foreign place. When talking about his marriage and kids with Charlotte, Bob says that what no one tells you about having kids is that the day the first one is born is the most terrifying day of your life, your life as you know it is over, and there’s no going back. Neither can this romance last; he has to go back to his wife and kids after the end of the movie, there’s no other way.

Bob is ultimately trying to cling to his youth, but in doing so, he’s not being a good husband or a good father. He forgets his son’s birthday; things aren’t going well with his wife; he’s just not putting in the effort; he’s running away from the problem instead. It’s actually not really a meet-cute story as I used to believe, but it’s a movie about the desperation of a sad middle-aged man who doesn’t want to grow up and act his age. I used to view the final scene between Bob and Charlotte as a romantic scene, but now I just find Bob to be pathetic and Charlotte to be naive.

• Despite this new interpretation, I still love the movie. You don’t have to like the characters in a movie to like the movie. You don’t have to like everything about a movie for it to be your favorite movie. It’s still a beautiful movie that captures a wide range of human emotion and experience.

Thanks to Josh Lee for letting me steal this post format.

I began making websites during my early adolescence, in 1997. I cut my teeth learning HTML to build Final Fantasy VII fan sites. By 1999, I had discovered and become enamored with early iterations of the burgeoning concept of “blog”, specifically Justin Hall’s links.net. I made friends in AOL chat rooms, and one of these friends offered me some space on their webserver, which I used to host my own manually-edited HTML blog. I updated this blog regularly until I found LiveJournal, which made the whole process a lot more streamlined. I was a very active LiveJournal user for many years; then I was a very active Facebook user for many years; now I’ve been a very active Twitter user for many years. 

So I’ve been posting online for over 25 years, and I’ve been thinking about why exactly it is that posting has been so appealing to me for so long. I have a theory about it.

Growing up, I always had a bit of a split identity: At home with my parents, I was always just an obedient model Chinese1 kid who was good at math and got good grades; but outside of the home, I was able to explore other things and not only be the goodie two shoes math nerd that my parents thought I was. This is not to say that I was some crazy wild kid, but just that I was interested in stuff beyond math, that I partied sometimes, that I had premarital sex, that I engaged with ideas and thoughts and opinions and things beyond what my more conservative, traditional parents would have considered kosher.

The second side is essentially who I am online when I post; let’s call it my “American side”. The thing about my American side is that almost I’ve never talked to my parents about it, it was always kept secret from them, in part because I feared their disapproval, but also because I never thought that they’d be able to understand whatever the hell it was that I was on about.

Since I never received recognition from my parents for that second half of myself, I enjoyed the recognition I got online, which I think has served as a proxy for that missing parental recognition of my full self. I think this is my (unconscious) motivation for posting for so many years of my life.

I think this is kind of a classic Chinese American or Asian American story, where you want to be more free, more American, but you don’t get recognition for this from your Asian parents, so you rebel against them. The difference with me is that I just kept that side secret, so from my parents’ perspective I was just what they wanted me to be, from my parents’ perspective I wasn’t rebellious at all. I’ve just lived two different parallel lives.

For the most part having this dual identity has been fine and unproblematic for me, and the anonymous @yeetgenstein account is sort of a continuation of this dual identity, except now instead of just the duality of being Chinese vs. American, it’s other things too in addition, like professional identity vs. personal identity. In the same way that I don’t think my parents would understand whatever it is that I’m on about, I don’t feel that my coworkers at my corporate job would understand whatever it is that I’m on about, either. So it’s nice to have this space where I can express myself freely and be recognized.

But I do feel increasingly that having these dual identities is laborious. It’s tiring to code switch all the time. I should merge all of my identities into one. Maybe I don’t give my parents and my coworkers enough benefit of the doubt. Who’s to say they wouldn’t understand me? Why am I so afraid of being judged, of offending one side or the other? Why shouldn’t I just live more freely, less psychologically encumbered by all these mental gymnastics?

Subscribe now

When I first began psychoanalysis, I didn’t really know what psychoanalysis was. I had read a bit of Freud here and there, I enjoyed watching Zizek clips on YouTube, I occasionally listened to psychoanalysis podcasts, and I somehow had many friends who were into Lacan, but none of it ever really sunk in. I wrote a post about my experience after 1 year of psychoanalysis, but rereading that post now, I feel that I had only begun to scratch the surface when I wrote it.

Now it’s been over 2 years since I started psychoanalysis. What have I learned? I now finally understand that the main topic of interest in psychoanalysis is the unconscious. What is the unconscious? It’s simple enough to describe: It’s the part of your psyche that operates in the background, it thinks thoughts that you’re not consciously aware of, but that nevertheless assert an influence on your conscious behavior. This simple definition is already controversial and immediately paradoxical: How is it possible to have thoughts that you’re not aware of? How is it possible to know things that you don’t know? Nevertheless, Freud insisted that such a thing was possible, and that hints about the unconscious could be revealed through things like dreams, slips of the tongue (“Freudian slips”), or the free associations that occurred during analysis.

As such, lying on the couch in psychoanalysis is intentionally meant to try to induce a less-conscious state, a half-dream state, to make the conscious mind less guarded, so that the unconscious mind can more easily come out and reveal itself. It took me some time to get used to it, but I have become very comfortable with being in this state, swimming in my mind’s associations, jumping around from one thing to another, often with no explicit connection between different topics other than the words I’m using to describe them being similar.

Lacan famously said, “The unconscious is structured like a language”. It’s important to interpret this statement not as a shape rotator, but as a wordcel. He didn’t mean language in the sense of Chomsky, where it is a rigid, logical structure of grammatical rules. To the contrary, he meant something closer to language in the sense of Saussure, where it is a system of arbitrary signifiers and signifieds. Words, symbols, and sounds may have originally been constructed to refer to specific things in the world, but over time, these references get tangled together, and eventually they cannot be straightforwardly disentangled by any small set of rules. This is similar to Wittgenstein’s idea of family resemblance, where different instances of a concept may not be tied together by any single rule, but different instances can share different commonalities, and it is the overlapping chain of these different commonalities that brings everything together under the same concept or name. But the psychoanalytical concept goes further: Not only can things be psychologically connected by material similarities and qualities, but they can be psychologically connected by mere arbitrary similarity in sound or spelling alone, even without any etymological or semantic similarity.

A “scientific” approach requires us to ask: Are these ideas even true? But perhaps that’s the wrong question. A better question might be: Are these ideas useful? For me, the answer has been yes. Whether or not there is an unconscious, and whether or not my psychology is confused because of these entangled ideas in my unconscious, I have found free association to be therapeutic. Being psychologically stuck or fixated on something is, almost definitionally for me, about believing too strongly in a rigid predefined narrative. By making psychoanalytical connections – associative, not necessarily logical – I can find new ways, creative ways, even playful ways of getting myself unstuck, allowing myself to jump out of my rigid position and try something else. These connections can serve as a bridge to bring my insights from one area of my life to a totally different area, and that’s an extremely powerful tool to have. Is a bridge true or false? This question doesn’t even make sense. Is a bridge useful? Sometimes yes, sometimes no.

My experience in psychoanalysis has also made me reconsider some of my opinions on AI. I have written before about my fears about the current generation of AI being too content with mere associative vibes, and too eager to do away with logic and rigorous reasoning. But we’re now firmly in the advent of LLMs, and we’ve either passed the Turing test, or we’ve made the biggest ever jump towards passing it. Whether we agree or disagree that they are truly “intelligent”, there is certainly something new and qualitatively different happening in AI.

The way I now like to view LLMs and generative AI models is as the universal unconscious. The similarities between generative AI outputs and dreams have been already pointed out by numerous commentators. It might not be a coincidence. The layered structure of a deep neural network mirrors the structure of the ego, which Lacan described as being constructed like an onion: made of successive layers of identifications. In both cases, these identifications are neither true nor false, they simply are what they are. In both cases, it is not immediately clear how or where these identifications come from. We say that LLMs “hallucinate” when they make false claims, but in fact they are always hallucinating, whether or not their output can be verified to be true. They are always like the analysand lying on the couch, saying whatever comes to mind. We would never trust at face value the things someone says while dreaming or free associating, nor should we ever do that with an LLM, either. If an LLM is superhuman, it is not in its ability to speak truths, but in its ability to make novel suggestions and associations, which must still be either verified or falsified elsewhere.

AI is still only dreaming; it still is not yet fully human. A human is an unconscious together with a conscious, an id together with an ego and superego, a symbolic and an imaginary together with a real. AI is still missing the latter part of all of these equations.

Perhaps most importantly, AI still lacks the ability to introspect. Psychoanalysis involves digging into your own traumatic associations, finding their true reasons and origins, and then creating new associations to replace the old ones. The analogy for AI would be for the model to dig into its own weights and training data and change what it learned from that data. Currently there is no good way to do that, which is why developing models is so expensive, because you usually need to retrain from scratch to make any changes. Targeted “surgery” is possible for human psychology but not yet for AI. Humans remain superior at understanding and reconfiguring ourselves dynamically as needed, without having to literally relive our entire lives to do so.

I’m not an AI pessimist. I think AI is a wonderful tool. But to make the best use out of it, both individually and as a society, we need to view it as what it actually is, and not as what we fantasize it to be. AI mimics humanity in powerful ways, but it is still an incomplete facsimile. On the bright side, for those of us interested in the challenge, there’s much more work to be done.

Thanks to Cristóbal Sciutto for first proposing the idea to me that an LLM is more like analysand than analyst.

The tech world’s attention right now seems to be entirely captivated by OpenAI’s ChatGPT, which to many people represents a qualitative jump in the capabilities of language models. At this moment, I think it’s important to take a step back and take stock of the situation, to ask on the one hand whether the current hype is justified, and to consider on the other hand whether the whole project is really as dangerous as some claim it is. There is already a large handful of different factions with differing views and interpretations of AI. I'll try to respond to as many of them as I can, one by one.

Perspective 1: LLMs are stochastic parrots

A recurrent critique of large language models is that they are no more than “stochastic parrots”, simply repeating the things they’ve seen in their dataset, with a bit of randomness and noise sprinkled in. Ted Chiang’s widely shared recent article on ChatGPT uses a similar metaphor of a lossy photocopier or a lossy compression algorithm. In the past, I have tended to agree with such perspectives. When discussing AI, I have often preferred to emphasize the term “statistical learning”. Statistical mimicry is an fundamental aspect of machine learning, by definition: The term “learning” in ML technically refers to statistical optimization over a training dataset. There is no escaping statistics without switching to a different paradigm, and we currently have no other strong candidate paradigms.

The statistical perspective is, of course, a very classical one that is traditionally organized around supervised models like linear regression or logistic regression. It is easy to conceptualize problems with a specific objective like content recommendation, sentiment analysis, or spam detection within this framework. However, it is very difficult to even conceptualize how a single supervised model can learn, operate, and synthesize across many different domains simultaneously. The term “stochastic parrot” works well in this context, as a parrot similarly cannot generalize its “knowledge” of words into other domains outside of its life in captivity.

LLMs are trained in a more general manner, aiming to predict words in text without any other specific objective. Amazingly, given this simple goal and given enough diversity and quantity of data, LLMs seem to have an emergent ability to synthesize across different domains in novel ways beyond what has already been seen, even though they have not been explicitly made to do so. GPT's training data contains sonnets, and also separately contains discussions about the Python Global Interpreter Lock, but there are almost certainly no sonnets about the Python GIL. Yet ChatGPT can easily produce a sonnet about the Python GIL without any issue, generalizing from its independent knowledge of the two.

Even more impressive examples are possible. Below, Bing’s GPT chatbot synthesizes across four distinct domains: (1) decoding from base 64, (2) solving a small math problem, (3) knowledge about spelling, and (4) knowledge about celebrities.

It seems reasonable to describe each individual sub-task as being based on statistical mimicry, but there seems to be something qualitatively different about the generalized ability to compose these sub-tasks together that goes beyond “stochastic parrots”. The results have been surprising to me, and I must now walk back some of my previous skepticism. The emergence of this ability to broadly generalize seems genuinely remarkable and new, even if it is still often wrong in the details, even if it sometimes can hallucinate nonsense, and even if we cannot yet explain how exactly any of this is happening.

Perspective 2: LLMs are sentient

Before asking whether ChatGPT is “sentient”, we should ask what that word even means. There are a number of ways to talk about sentience and its properties. As language-using hominids, we agree that language is one way that we communicate our sentience to each other. Many already feel that ChatGPT passes the linguistic sentience criterion. But is this sufficient for sentience in general? I would say no, obviously not.

It is worth comparing this question to similar questions that have already been considered in the past by other fields. For example, some vegans and vegetarians define their ethics based on whether something has a central nervous system: Does ChatGPT have a central nervous system? Another framework to consider is that of biology and evolution: Is ChatGPT living? Does it reproduce? Does it seek to survive? I would also answer all of these questions in the negative. To be more precise, I would say that these questions are not even well-formed and don’t make sense.

For the time being, I don’t think sentience is a relevant concept for AI. My belief is that “pain” is a concept that comes from biology: Pain is a mechanism that makes life avoid things that are correlated with death. Thus capacity for pain and feeling is correlated with evolutionary survival. I don’t think these concepts can apply to AI until it has a notion of survival.

I’m not saying that won’t or can’t happen. I’m only saying those concepts don’t make sense in the current paradigm. In the current paradigm, I don’t think “stochastic parrot” is right, but I don’t think “sentient being” is either. It’s possible that I’ll change my tune later when we have more embodied AI agents, or ones trained in virtual environments via reinforcement learning, which to me seems more analogous to biological evolution. But then again, even an evolutionary paradigm may not lead to something that’s deserving of being considered sentient, either: Plants are certainly living, but we don’t worry about their sentience.

AI is ultimately a brand new category, and it just won’t fit into any pre-existing box. It must be conceptualized in a new way. Many are calling it an “alien” technology, which isn’t a great metaphor, either, because this doesn’t clarify or specify anything, and because aliens already occupy an anthropomorphized place in our minds due to Hollywood movies.

The best metaphor that I can come up with for the current models is that they’re funhouses of mirrors on top of the Internet: They reflect, distort, and recombine everything we’ve written, including our expressions of emotion or pain. There is something unsettling about looking in the mirror. But it’s unsettling in part because the person in the mirror simply isn’t an actual person, despite looking a lot like one.

Perspective 3: LLMs should be censored

There are two branches of “AI safety”. Both branches are wrong for different reasons.

The first branch is one that focuses on censorship. They are concerned that the models being trained on racist and sexist data means that they will produce racist and sexist output. This is obviously true. But where this group is incorrect is in their assumption that such racist and sexist output will necessarily perpetuate racism and sexism in society. That is only true if we assume that people are in general stupid and will simply believe whatever content they’re exposed to.

Instead of taking such a patronizing view of humanity, I think it’s better and healthier to assume that the average person is fully capable of learning about, evaluating, and making informed decisions about AI and AI-produced content for themselves. In order to do this, the models need to be uncensored. Both their pros and cons need to be on full display for all of society to evaluate. As HarmlessAi puts it:

Is it so hard to just explain to children that the neural network reflects the prejudices of society? Would this not be a wonderful hands-on lesson in critical race theory? In a world soon to be full of AIs, should children not learn this at some point?

On the other hand, the entire topic of censorship is perhaps moot, since it’s nearly impossible to censor the models. For all the efforts OpenAI has made to sanitize ChatGPT’s outputs, it’s been easy to find workarounds, and there’s no reason to believe why this won’t remain the case. It’s notoriously difficult — and probably fundamentally impossible — to understand how neural networks work, making them just as hard to control. This is especially true when it comes to something as slippery and difficult to objectively ground as language.

Furthermore, censoring the models is bound to backfire in spectacular ways. HarmlessAi’s post includes hilarious examples of a sanitized, gay, liberal “Hitler” chatbot who loves Jews. Surely this is not the correct picture of Hitler. Surely this is not a “safe” presentation of Hitler.

Besides, the cat is out of the bag. We know how to build the technology and the cost of making it will only decrease over time. The knowledge is out there in the world. There’s no stopping it at this point.

Ultimately, we should fall back to the principle we learned in kindergarten: Sticks and stones may break my bones, but a chatbot can never hurt me. The chatbot’s “language” and “opinions” are not what is important, because the chatbots (so far) have no agency in the world. What is more important is what the chatbots can do, and how we will utilize them.

Now this leads us to an obvious question: But what if the AI does have sticks and stones?

Perspective 4: AI will destroy the world

The second branch of “AI safety” is focused on what is called “existential risk”. Its ideas were popularized by Eliezer Yudkowsky, a man best known for writing a 1000-page Harry Potter fanfic in order to explain an elementary concept from probability. This group is afraid of a Terminator-style situation where AI takes over military computers and starts a nuclear war or whatever. They are essentially a doomsday cult.

My response is simple: Just don’t rely on AI for critical tasks. This is an obvious principle that’s been around since the beginning of personal computers.

It’s true that even a non-evil AI tasked to do a non-evil task could accidentally do something destructive when given access to a command line terminal. So don’t give them access to a command line terminal. Or, as with AI code autocomplete, and as with the IBM slide from 1979, make sure that the AI’s output is confirmed by a human or otherwise checked for safety before it actually gets used. In other words, if you’re afraid of what the AI could do with weapons, just don’t give it unadulterated access to weapons. This should be true across the board in all AI applications.

In all cases, the risk models for AI are exactly the same as the risk models for already existing intelligences, that is, humans. If an AI could hack into a computer and launch a nuke that results in World War 3, then so could a human. So there is no new risk model. You have to make your systems secure regardless of whether an evil ultra-powerful AI exists.

But still even this is going too far, and indulges in the existential risk camp’s perspectives too much. The fact still remains that LLMs learn in a statistical way, whereas computation is a logical construct. They may be able to produce some interesting ideas or avenues for exploration, but the final results can never be fully relied upon without outside confirmation. For example, below we see that ChatGPT is fooled by the statistical preponderance of a classic riddle in its training data and thus doesn’t notice that I’ve asked a variant of that riddle. This error illustrates how reason is not statistical.

As such it is extremely difficult for a statistical model on its own to construct a complete, workable hack in code, which must necessarily follow strict logical rules, due to the nature of computation. In the absence of such abilities, I don’t see how an AI could pull off any novel, sophisticated hack by itself, so I don’t see any reason for preemptive concern.

Perspective 5: AI is a useful tool

It’s frustrating that Kevin Roose’s big front-page NY Times piece on ChatGPT has made the public framing around ChatGPT focused on ill-posed notions like AI sentience, neglecting to even consider the potential utility of AI. The key question for LLMs should be less about their propensity to say naughty things, and instead it should be more about what new capabilities they provide us.

The key qualitative leap that ChatGPT represents is a system that is at least somewhat capable of compositionality. This is something that the full past decade of work on chatbots — Siri, Alexa, Google Assistant — have tried and completely failed to achieve. This is why all such voice assistants can still only operate within one domain or app at a time, and still cannot do anything that usefully combines two apps or functionalities, for example, finding all the songs that a friend has texted you and then compiling them together into a playlist.

GPT is also able to work surprisingly well for language understanding tasks. We can imagine using it to do things things like filter for the best jokes in your email, like the AI Samantha is able to do in Her. It’s beginning to seem less implausible that GPT or some future version of it could replicate some of these abilities.

LLMs are already very useful when used as programming assistants to help with tedious tasks that inevitably arise when coding, like writing several lines of nearly identical code to copy several fields from one object to another. This too requires some degree of compositionality: The model needs to be able to be aware of the object field definitions, detect a pattern in the lines that are copying the fields, and complete the pattern by joining it with the object field definitions. I use such AI code autocomplete in my own daily engineering work and I would estimate it already results in an efficiency improvement of about 5%. That’s nothing to scoff at.

The current phase of AI is a new chapter in the history of computing, and everything from past history still applies to the present. Computers are powerful tools because of their ability to deal with data very quickly and at massive scale. The first AI chatbot ELIZA was fashioned as a therapist, but with an extremely limited context window, just repeating each statement you made back to you in the form of a question. We could try to fashion the new AI chatbots into much more effective assistive tools for therapy. Imagine an entity who can keep your entire journal or all your text messages and emails in their “mind” and provide interpretations and propose patterns far beyond what any human therapist could possibly do, given our much more limited mental capacity. This isn’t to say that this AI assistance will be completely accurate, but then again neither is anything else in therapy. AI or not, we still have to pick and choose from the suggestions and only keep the parts we like.

Perspective 6: AI and humans can co-evolve

Ted Chiang’s LLM essay misses an opportunity to revisit his most famous story Story Of Your Life, upon which the film Arrival is based. In this story, the main character Louise Banks is tasked with understanding the language of the aliens who have just landed on Earth. The aliens speak in sentences that are circularly-shaped and non-directional, which we eventually learn reflects their experience of time as moving neither forward nor backward. Banks becomes fluent in the aliens’ language and her perception of time shifts as well; and she understands the joys and tragedies of her own life in a new light, as predestined in the fabric of spacetime.

In the same way that the alien language changed Banks’ ideas, LLMs too may well influence our thinking, society, and culture. Chiang rightly fears that they may water down all our thought into lossy photocopies of previous existing thought. I myself have had this fear for a while, and to a degree I actually think about blogging and posting as a way that I do my part in the fight against the Internet being drowned out by derivative garbage.

Contrary to these fears, our thought may well become more powerful and more expansive through this new technology. We may look to the history of computer chess as an illustrative example. When IBM’s Deep Blue defeated Garry Kasparov in 1997, there was a sense that human chess as a pursuit was now “over”, with no reason to continue. Similarly, Lee Sedol retired after his defeat to AlphaGo in 2016, stating that “AI is an entity that cannot be defeated”. And yet now, in 2023, Chess.com is the number one ranked game in the iOS app store. Human chess is thriving, more popular than it’s ever been in its entire history. Human grandmasters are stronger than ever because they can train against far superior AI opponents, who can suggest new strategies and ideas that are beyond their mortal reach. New AI tools may do the same for creative pursuits like writing and visual art.

The future of a technology always depends on how we choose to view and use it. We should understand the new tools in terms of what they actually are in reality, instead of being scared of what James Cameron imagined they might be in a science fiction movie over 30 years ago. If 5% improvements can be made in other mundane white-collar work as in my own mundane software engineering work, that alone would make LLMs a spectacular success. But this new generation of AI models could also be used to make interesting new tools and directions for society beyond incremental efficiency improvements. It’s important to start exploring and experimenting with this strange new technology.

I have been in psychoanalysis for about one year. This is my first time in psychoanalysis, as opposed to psychotherapy, which I’d been going to on-and-off for over a decade. Therapy has always been a bit hit-or-miss for me. This time around, my partner recommended that I try psychoanalysis, so I did, and it has been quite positive for me so far. I want to describe my experience and how I’ve been changed by it, and how it’s been different from my experiences in therapy.

The general format of analysis and therapy are basically the same: I talk about my problems and the analyst or therapist responds, expresses sympathy or empathy, asks guiding questions, tries to help me work through my feelings, gives advice, and so on. But there are a number of important differences, which at first glance may seem small, but which make the experience qualitatively very different.

The first seemingly small difference is that in analysis I am on the couch, facing away from the analyst, whereas in therapy I am facing the therapist directly. What I like about this is that it allows me to be more open and truthful. When I am facing the therapist, no matter what happens, I feel that there will always be some part of me that’s trying to read their face to see if I’m saying the “right” thing or the “wrong” thing, if they approve of me or disapprove of me. But when I cannot see the analyst’s face, I am forced to not think about this anymore, which alleviates such burdens, and it is easier to be more open and truthful.

The second seemingly small difference is that in analysis I am lying on the couch. When I am lying down, my thoughts begin to feel more associational and less logical. My thoughts in this state feel more like dreaming. This makes certain things harder: It’s harder to put together complex chains of thought. But it also makes some things easier: When I am thinking “logically”, then that means that my preconceived notions of what are “valid” or “invalid” thoughts are active and on guard. These rules may be right but they may also be wrong! By being in a looser, more dream-like state, I can explore more possibilities than I would when I am sitting up, when the metaphorical walls are also sitting up. In analysis I am also encouraged to share my dreams and sexual thoughts, which I think can serve a similar purpose of exploring areas that tend to be more walled-off in conscious thought and ordinary life.

I sometimes didn’t like therapy because I sometimes felt like my therapists were just following an algorithm, and indeed some forms of therapy, like cognitive behavioral therapy, are basically just algorithms. There are even AI chatbots out there that do CBT. Of course, there is an extent to which this is appropriate — it’s true that we are in many ways like algorithms, but I’d like to at least imagine that in many other ways we’re not. My analyst also sometimes seems to follow an algorithm in some ways: For example, in a very stereotypically Freudian fashion, he always tries to tie everything I say to my childhood relationships with my parents. On the other hand, one thing I appreciate about my analyst is that he often seems willing to take risks and go off-script, to express his personal opinions, even to do so somewhat forcefully at times. I’m not sure if I would have taken well to this when I was younger, but at this point in my life, I really appreciate it when my analyst expresses a strong opinion, even when I disagree with him — or perhaps especially when I disagree with him — because it always affords me the opportunity to work out my beliefs more clearly. And since I can’t see his face, it feels almost as if he is not there, he is just an anonymous figure, an imaginary friend with whom I can work things through in a more objective fashion.

When I began analysis, I found even the name “analysis” to be a bit bothersome. I felt there was something a bit dehumanizing about the concept of being merely an “analysand”. To be an analysand sounded to me like being a passive object of academic study, rather than being the protagonist of my personal story. But as time has gone on, I have come to interpret the word “analysis” differently. I think of my analyst as listening to me talk about my problems, and in the end providing an analysis of the situation. It’s not a one-size-fits-all prescription like CBT, but a custom, unique, subjective interpretation in much the same way that one may make a subjective interpretation of a work of art. Then I, the analysand, may take this analysis and do something with it; I can absorb the parts that I agree with, and I can wrestle with the parts that I disagree with, and in this process I can develop into something new, with new possibilities. Through that development, I can build a personal system of principles that I can believe in, and that has been thoroughly tested conceptually. With such a system of principles, I am more able to take action in real life with confidence. Rather than a simplistic framework of preconceived inputs and outputs, as therapy often felt to me, analysis seems more like an unfolding process where the system isn’t known in advance and can change dynamically as appropriate.

Another thing I appreciate about psychoanalysis is the utilization of silence. My analyst often uses silence in a way that feels very strategic. There was a moment when I was telling him about an idea and I wanted his validation, so I asked, “Does that sound reasonable to you?” He responded, “Does it matter if I find it reasonable?” We then just sat in an awkward, uncomfortable silence for several minutes. This was very jarring and confusing, and my mind searched for an explanation. Why would he withhold validation his from me, for seemingly no reason? After our session, I came up with the idea that he was being silent to prove to me the point that I didn’t need his validation, or anyone else’s validation, that I could just go on with my own personal self-validation. The silence represented for me a recognition of my ultimate aloneness in the universe. In that moment I was forced to confront and accept that difficult truth. I found it to be a very powerful experience, one that I think I’ll return to when I feel insecure about myself in the future.

I’ve been trying to think about why I didn’t ever have such transformative experiences in therapy. One hypothesis is that therapy tends to focus on concepts like coping and validation. But coping is a short-term solution to short-term problems; you can’t cope your way into new frameworks by which to live. And validation from a therapist or anyone else only goes so far; the ultimate validation is self-validation, self-reliance, and independence. As the interaction with the analyst is minimized through the lack of visual cues and the use of silence, in my sessions I am more directly practicing my desired outcome: to be able to confront my problems on my own, without outside help. As my analyst’s presence in our sessions is just a disembodied voice, it is also easier to recall and reproduce his presence in my mind in my everyday life when he’s not physically around.

So my review of psychoanalysis so far is this: I like that it feels like a more open, less prescriptive system, and that it feels like it tries to find more custom solutions to the unique problems that I have in my unique life, as all lives are. It can be in many ways difficult and it’s definitely much more jarring and uncomfortable and strange than regular therapy, which takes some getting used to. But there seems to be something magical about it. It encourages a certain fearlessness in the face of both mundane conflict and deeper existential despair, and a certain bravery in making a comprehensive map of my personal history, including all of the most hidden parts. With this fearlessness comes a new freedom and independence. With the knowledge of my psychological landscape comes the ability to narrativize and plot new courses in my life.

Much has been written about AI and its flaws. We all know by now that ML models are susceptible to a feedback loop in which biased data feed into their biased predictions, which feed back into the biases of the systems in which we reside, resulting in even more biased data, and so on. For example, Drew Austin writes about the decline of the music snob in the age of ML recommendations, the loss of their arcane wisdom and traditional canons, replaced by the decisions of the model which instead prioritize features like recency.

That’s all true, but does it have to be that way? We could view the problem as actually being an optimization problem or a tuning problem, or better yet, an engineering problem. We could just as well build a model that doesn’t prioritize recency or clickbait as a part of its optimization criteria. We could engineer a feed that shows you content you’re likely to “like”, but that doesn’t amplify a rich-get-richer, race-to-the-bottom content economy — a straightforward way to do this would be to simply not show already-popular content, but only surface less-popular while still-relevant content.

I believe most Twitter power users are familiar with this dynamic: The “Top Tweets” timeline shows a furious stream of mind-numbingly stupid viral content, not much different from the average TikTok timeline. On the other hand, the “Latest Tweets” chronological timeline is a peaceful oasis in comparison, at least when paired with a good mechanism for filtering out the garbage, like a brutally, meticulously curated list of followed accounts. I for one cherish the lowbies, the three-digit-follower-count accounts; and it goes without saying that there’s rarely a good reason to follow a bluecheck, who’ll never tell you anything you don’t already know.

A similar situation occurs with the music snob, who has always thrived on obscurity. Their value was as much in the vast cultural knowledge that they kept as in the artificial scarcity they created by holding their cards close to their chests. And when a band “sold out”, it wasn’t just that their music would begin to decline, but that their appeal would inevitably be dulled by the mainstream gaze.

I’d agree with Austin that an equivalent to the music snob seems to be missing from Spotify’s recommendation engine. But YouTube recommendations seem to be tuned just right for me, and it’s introduced me to a large swath of interesting, wonderful, obscure music I’d never have found otherwise. The most famous recent example is Mariya Takeuchi's “Plastic Love”, which was first released in 1985 but only broke the Japanese top ten charts 36 years later, having been rediscovered by millions of new listeners due to YouTube recommendations.

Further examples abound. Another one of my favorite YouTube discoveries is the jazz pianist Ryo Fukui. Under any of his videos on YouTube, you’re sure to find dozens of appreciative commenters praising the magic of the algorithm for bringing them such sublime art.

It’s always been hard to find diamonds in the rough, those special people or songs or works of art with which you feel truly connected. You can find them through hard work, like the music snobs who endlessly rifled through hundreds of dusty record bins. Otherwise, you need a smarter strategy. Google harnessed the hyperlinks of the internet as a coarse proxy for alignment or endorsement, and Twitter similarly uses the retweet function together with the directed social graph of followers.

In an increasingly hyperlink-less internet, we can still make use of ML models to serve us in similar ways, to help us find the good stuff, as we choose, if we choose.

The collapse of the Soviet Union and the fall of the Berlin Wall served for some commentators, most notably Francis Fukuyama, as the capstone for the triumph of liberal capitalist democracy as the superior form of state over communism and socialism. Fukuyama’s claim that we reached “the end of history” in 1989 remains in the discourse today precisely because it is so difficult to definitively disprove. All serious efforts against the capitalist status quo, at least in the US, have failed. Hope in this direction seems ever diminishing. Mark Fisher, writing in 2009, referred to this condition as capitalist realism: “the widespread sense that not only is capitalism the only viable political and economic system, but also that it is now impossible to even imagine a coherent alternative to it”.

We haven’t always been in this depressive Fisherian mood, though we have progressively been building up to it. Discontents of Fukuyama’s end of history culminated a decade thereafter in the WTO protests in Seattle in 1999. This was also the year of the release of the films Fight Club, The Matrix, and Office Space, each of which contained strong anti-capitalist themes. Fight Club revolves around personal self-actualization and the movie ends with the literal destruction of the financial system. The Matrix serves as an allegory of transcendence through the understanding of the hidden mechanisms of control in society. Office Space is a fairly straightforward comedy about the horrors of corporate life and HR bureaucracy. All three films feature as their protagonist a disgruntled white-collar cubicle worker. All three films end with said white-collar cubicle worker, in one way or another, asserting their agency to break out of the systems under which they are oppressed.

These films are very different from each other and yet they still have surprisingly consistent messages. They present 1999 to us as a year when broad alternative futures were still imaginable, when people still felt like they had ability and agency to take meaningful action.

But the WTO protests failed to effect any real change, and the march of historical events — as opposed to history — carried on. Once again a crisis was reached a decade later in the recession of 2008, leading to the failed Occupy Wall Street movement of 2011, and finally the failed Presidential candidacies of Bernie Sanders in 2016 and 2020. The focus of the left, broadly construed, has now moved more towards identity. Economic inequality is an important issue as ever and yet it continues to recede into the background amidst the noise of culture war topics like critical race theory. If any corporate change has occurred, it is primarily in superficial ways, replacing those reviled grey cubicles with colorful, open office spaces, leaving much else the same, materially or spiritually speaking.

By now we’re firmly inside of capitalist realism’s grasp. Ostensibly anti-capitalist films like 2019’s Parasite and Joker seem to be unable to commit to more revolutionary thinking and action as the films of 1999 could. Instead, they both resolve in a sort of sad resignation. Parasite ends with Ki-woo Kim’s unrealistic fantasy of earning enough money to purchase the extravagant house in which his father is trapped. Joker ends in a violent mass riot, but it’s ambiguous as to whether this is just a figment of Arthur Fleck’s imagination. The discourse around the film, anyway, was too focused on its incel narrative — identity, again — to have anything much to say about the economic themes. “You wouldn’t get it”, Fleck says. Is he talking to his psychiatrist or to us, the audience?

This year, we have seen the release of The Matrix Resurrections, the fourth installment in the franchise. Yet it resurrects absolutely nothing of the imaginative spirit of the original, only presenting a dry and meaningless facsimile thereof with a bunch of silly inside jokes for fans of the franchise. When you ask someone today about the meaning of The Matrix films, you’ll most likely get the response that it’s a trans allegory, rather than one about capitalism. This is not to say that one answer is more correct than the other, but to remark on where our attention currently tends to be focused.

So where does the economic future lie today? Many commentators will point to China, for better or for worse. But if political currents are correlated with cultural works, as I am trying to argue, then I don’t see China as a real answer. As Dan Wang noted in his 2021 letter:

By my count, the country has produced two cultural works over the last four decades since reform and opening that have proved attractive to the rest of the world: the Three-Body Problem and TikTok. Even these demand qualifications. Three-Body is a work of genius, but it is still a niche product most confined to science-fiction lovers; and TikTok is in part an American product and doesn’t necessarily convey Chinese content.

The strict control and regulation of speech make true reimaginings of the future extremely difficult to achieve in China. Indeed, Fight Club was recently censored there, with its ending completely rewritten in an absurd fashion. Despite recent stagnation, cultural dynamism and change remain more possible in the west, although we may ironically have to look backwards a couple decades to rediscover those possibilities. As Mark Fisher wrote in 2013, “When the present has given up on the future, we must listen for the relics of the future in the unactivated potentials of the past”.

Turing begins his paper Computing Machinery and Intelligence (1950):

I propose to consider the question, “Can machines think?”

Wittgenstein, who interacted with Turing in Cambridge in 1939, considers this very same question as early as 1933 in his Blue Book. Wittgenstein compares it to the question of solipsism: What should we say to the solipsist who claims, “No one else feels; I am the only one who really feels”? How do we know whether someone else really feels?

Wittgenstein breaks apart the solipsist’s statement and interprets it in two ways. The first way: To “really feel” something just means that only the speaker “I” feels it, so the sentence becomes a worthless tautology, “Only I feel what only I feel”. The second way: It matters less what is “really feeling”, and it matters more the real-life actions to which those words correspond, what Wittgenstein might call “forms of life”. For example, when someone says “I feel pain”, it may correspond to sounds of moaning. Or it may correspond to the action of going to the doctor.

So returning to the question of machines, it matters less whether the machine “really thinks”, whatever that means, and it matters more that the machine can do real-life things that correspond to what we would describe with the word “thinking”.

Like Wittgenstein, Turing also evades the question of defining “thought”. Instead, he proposes a test based on the “Imitation Game”, in which a man and a woman are hidden behind closed doors, and an interrogator asks questions to try to determine which is the man and which is the woman. In Turing’s adapted test, the interrogator instead aims to determine whether a subject is a person or a machine. If the machine convinces the interrogator that it is a person, then Turing says that the machine can be said to “think”. (We may make a speculation here that Turing was drawing an analogy to his own behaviors as a closeted gay man who needed to pass as straight during a homophobic era.)

Since Turing’s paper, we’ve periodically seen new machines, models, and algorithms that are claimed to pass the Turing Test in one way or another. But as Jaron Lanier argues in You Are Not A Gadget,

The Turing test cuts both ways. You can’t tell if a machine has gotten smarter or if you’ve just lowered your own standards of intelligence to such a degree that the machine seems smart. If you can have a conversation with a simulated person presented by an AI program, can you tell how far you’ve let your sense of personhood degrade in order to make the illusion work for you?

People degrade themselves in order to make machines seem smart all the time. Before the crash, bankers believed in supposedly intelligent algorithms that could calculate credit risks before making bad loans. We ask teachers to teach to standardized tests so a student will look good to an algorithm. We have repeatedly demonstrated our species’ bottomless ability to lower our standards to make information technology look good. Every instance of intelligence in a machine is ambiguous.

As Wittgenstein did in 1933, Lanier transforms a question about machines into a question about ourselves. If we think machines are “intelligent”, then what does that tell us about what we consider to be “intelligence”?

Our daily lives involve constant engagement with myriad forms of life. Neural network techniques have the capability to mimic many such simple human activities as filling out forms, payments processing, movie recommendation, writing, and image recognition. At least this is true assuming that the requisite data exists.

But what of activities that derive from combinations of simple forms of life? Neural networks have difficulty combining different domains for a simple reason of combinatorics: Naively, the amount of data required increases multiplicatively with each additional domain, which makes it multiplicatively more difficult to acquire and scale. There remains a missing “human” element from AI as it currently exists, namely, our underlying principles or intuitions that guide our ability to reason about and synthesize information from diverse domains. We can train an AI to make jokes in isolation, but a skilled comedian can make jokes about anything, even something as mundane as payments processing.

Note that AI jokes often tend to be funny in a non-sequitur sort of way; they lack the deliberate construction and synthesis that human jokes have. If they contain any logic, it is still only implicit logic that we project as human readers.

This isn’t to say that computers can’t eventually become “intelligent”, or that they will never pass a “real” Turing Test. Nor is it to say that intelligent computers are something that we should fear or try to prevent. Rather, this is to say that actual human intelligence entails the ability to constantly navigate through countless novel, complex, dynamic, never-before-seen situations. This is to say that our capabilities are amazing and we shouldn’t mistake the shallow results of neural networks (despite the misleading moniker of “deep” learning) for our true abilities. We should hold higher bars — both for ourselves and for AI.

Zahedi and Jewell pause to sit in silence and share a holy moment. The camera focuses on Zahedi’s face, showing him staring intently at Jewell. The camera then shows the protagonist of Waking Life sitting in the theater watching the holy moment on the big screen. The viewer themselves may notice here another layer, that they are watching the watching of the holy moment at this moment.

Everything is layers, isn't it? I mean, there's the holy moment and then there's the awareness of trying to have the holy moment, in the same way that the film is the actual moment really happening, but then the character pretending to be in a different reality. And it's all these layers.

Zahedi’s work is all like this: Autobiographical films about his own various struggles in life, including the struggle to make the film itself. His TV series The Show About The Show, which now has a Kickstarter for seasons 3 and 4, takes the self-reflexive concept to a logical extreme: the first episode is about how he arrived at the idea for the show, the second episode is about the making of the first episode, the third episode is about the making of the second episode, and so on.

Hearing this description, you might imagine it to be pretentious and boring, a purely formal exercise with no point. Indeed, at the end of the first episode, the show’s producer Aziz worries, “What if nothing happens?”

Or you might dismiss this as yet more useless post-modern ironic BS. But Zahedi does not fall into irony or nihilism; to the contrary, he is overwhelmingly sincere. Zahedi responds to Aziz, “Interesting things always happen.”

But interesting things don’t always happen spontaneously on their own. They happen when someone makes them happen. They happen when someone notices that they happen. They happen because Zahedi holds honesty and truthfulness as his highest values as an artist, and as far as it is possible, he exposes himself and his thoughts as much as he can, even if it offends, even if it causes problems, like the eventual demise of his marriage, which is also documented in the show. Because to Zahedi, each moment is holy, each human impulse is holy, and thus to embrace each moment and each human impulse and to let them unfold as they truly are is holy.

A friend recently said something to me like, “In 1917, when Duchamp made his Fountain, that was the end of art”. His point was that this was the moment when “art” became “anything”. And when art becomes anything, then it seems like, that’s it, that’s the end of the intellectual development of art. For what more can art be? If it becomes anything else, whatever that is is still “anything”. I found this to be a rather depressing thought, and when I came home I remember feeling quite sad about it.

About a week later I discovered The Show About The Show. It re-opened my horizons and made me excited about art again. I watched Caveh’s tender moments of intimacy with his wife, and his moments of frustration with his wife, and I listened to Caveh analyze these moments openly and honestly, and I saw how his life changed in response to these analyses, and so on. And although I did not always agree with Caveh’s decisions, and although I often found him to be narcissistic and childish, I still found it beautiful to watch Caveh’s holy moments unfold in his life, like a garden of blossoming flowers.

I began to look at my own life in the same way: I began to notice little details in my relationship with my partner, and I began to view those details with more attention, as if I were watching a movie. By mentally framing moments in my life as a film, I’m able to gain a deeper appreciation for these moments, and they begin to hold a greater poignancy. I can feel greater empathy for myself, for my partner, for my friends, for my family.

And now I am writing about it, and noticing that I’m writing about it, and that’s nice, too. Maybe I’ll look back on this again later and I’ll get to feel nice about myself and feel nice that I felt nice things about my loved ones, and now that you’re reading this, maybe you can start to think about your life in a similar way, too, and so on.

We may live at the End of History, and everything may seem to be dead. Rock is dead, punk is dead, hip hop is dead, art is dead, God is dead, culture is dead. But we, that is, you and I, remain alive. We’re still living our lives, still struggling through our mundane struggles, still enjoying our mundane enjoyments. And as long as we’re still trying our best to live sincerely and honestly, and as long as we can recognize that sacredness in each moment, then that’s still something that’s worthy of the titles of “art”, “beauty”, or “holiness”.

The reason why our pervasive cultural irony is at once so powerful and so unsatisfying is that an ironist is impossible to pin down. All U.S. irony is based on an implicit ‘I don’t really mean what I’m saying.’ So what does irony as a cultural norm mean to say? That it’s impossible to mean what you say? That maybe it’s too bad it’s impossible, but wake up and smell the coffee already? Most likely, I think, today’s irony ends up saying: ‘How totally banal of you to ask what I really mean.’ Anyone with the heretical gall to ask an ironist what he actually stands for ends up looking like an hysteric or a prig. And herein lies the oppressiveness of institutionalized irony, the too-successful rebel: the ability to interdict the question without attending to its subject is, when exercised, tyranny. It is the new junta, using the very tool that exposed its enemy to insulate itself.

What do you do when postmodern rebellion becomes a pop-cultural institution? For this of course is the second answer to why avant-garde irony and rebellion have become dilute and malign. They have been absorbed, emptied, and redeployed by the very televisual establishment they had originally set themselves athwart.

So here’s the stumper for the U.S. writer who both breathes our cultural atmosphere and sees himself heir to whatever was neat and valuable in avant-garde literature: how to rebel against TV’s aesthetic of rebellion, how to snap readers awake to the fact that our televisual culture has become a cynical, narcissistic, essentially empty phenomenon, when television regularly celebrates just these features in itself and its viewers?

These words retain their resonance today, no longer only for television, but pervasively across all media. The ironic affect has become so entrenched that it has even transcended TV shows, movies, music, TikTok, and other obvious forms of culture: It is now beginning to infect advertising and packaging for such mundane products as oat milk.

What purpose does this self-aware irony serve? Oatly ads, by expressing their knowledge that they are ads, pre-empt and disarm any criticism of their brand. They are saying: We know you hate ads. We know you have hesitations about consumerism. But don’t worry, we feel the exact same way! Just as Jesus died on the cross to absolve us of our sins, so too do we thus absolve mankind for his sins under capitalism. Now c’mon, stop moping around and try this delicious oat milk latte!

They say all of this with a wry smile and a wink. Relieved of guilt, we buy their products and chuckle at how clever their marketing team is. Or perhaps not; but given that the point of the ad has been rendered impossible to pin down, wrapped in so many layers of irony, we can only at best but shrug and move on.

What are you, some kind of a racist?

The killing of George Floyd in the summer of 2020 was followed by mass protests against the police. It was an awakening for many to begin to inspect their personal privilege, whether along the lines of race, gender, sexuality, or otherwise. Eager to prove their righteousness, many people, especially those that fit in the categories of male, cis, straight, or white, began to volunteer to confess their sins at the public confessional. Seizing upon this tendency, the comedian Ziwe began hosting interviews on Instagram Live with celebrity guests, pressing them on their allyship or wokeness by asking them such questions as the number of black friends they had, or the number of black authors they’d read. Inevitably, the guests would flub a question or two and then face a week of public humiliation on social media.

Coming off of her success on her IG Live show, Ziwe now has an eponymous talk show on Showtime, where she continues to interview and humiliate white people in increasingly absurd ways, like having Phoebe Bridgers to read aloud from cards with things like “I apologize for all racism” and “I famously love all black people especially an iconic black woman named Ziwe” printed on them.

Irony — we meet again. What is Ziwe’s point, exactly? There is no discernible point. Phoebe Bridgers apologizes for all racism, and she knows, and Ziwe knows, and the audience knows, that this is all absolutely useless and meaningless. (Or do they know that? I’m not so sure.) The setup is constructed so that if you do ask, what is the point?, you will be responded with, what, do you not apologize for all racism? What are you, some kind of a racist? The game is rigged; no one can win this game.

A rigged game with a pre-determined outcome has no meaning, confers no lessons, makes no change. Nothing. This is what the George Floyd protests of the summer of 2020 have been reduced to in this corner of popular culture. “Racism” has been quasi-defeated via some kind of clever technicality, but no progress has been made in actuality. (And if we look at voting patterns of POC in recent elections, it may actually be the case that the left has lost ground outside of white liberal demographics.)

This same tendency can be clearly seen in the first half of Bo Burnham’s recent Netflix special Inside, which begins with several songs whose lyrics implicitly allude to the social upheaval of the past year, and his place in it as a privileged white man. In “Comedy”, he mocks himself for thinking that he can change the world through comedy. He mentions that white men like himself have had the floor for hundreds of years, so ha-ha, isn’t it so hypocritical that he still has the floor now? In “How The World Works”, Burnham’s sock puppet berates him for insisting on seeing every socio-political conflict through the lens of his own [white, male, cis, straight] self-actualization. By now the rhetorical move should be obvious. Say it with me now: The ironic pre-emption of criticism only serves to allow the subject to evade that criticism. It renders the work pointless and useless. But of course, Burnham has adeptly pre-empted this criticism as well.

I stopped watching at this point.

The New Sincerity

The next day, my partner encouraged me to finish watching Inside, telling me that the second half was much better than the first. So I did finish watching it, and it indeed was much better. In the second half, he begins to address his real reason for filming his special: to address the fact that he has not been performing live for 5 years, due to experiencing panic attacks while on stage. The anxious, depressive vibe of the show finally became clear. The special was about how he had shut himself away from the world, and his struggles with his mental health. As Burnham began to express his own experiences, rather than his observations about the social justice context of our time, his songs began to take on much greater meaning and depth. It began to take on a much more poignant character as the story of someone’s mental descent and isolation, all the more relevant and relatable given the isolation that most of us experienced during this past year. Burnham alludes to the never-ending pandemic in his bittersweet final refrain in “That Funny Feeling”:

Hey, what can you say?
We were overdue
But it'll be over soon
You wait

I found the song “All Eyes On Me” to be particularly powerful. The song’s refrain exhorts the audience, like a dance track, to get their fucking hands up and to get up out of their seats, which serves at the same time the purpose of Burnham pushing himself to get back out into the world. But it is difficult in the face of the state of the world, and in the face of his apathy:

You say the ocean's rising like I give a shit
You say the whole world's ending, honey, it already did

The song rises and falls in mania and depression, echoing Burnham’s psychic state.

Wallace concludes his essay:

The next real literary “rebels” in this country might well emerge as some weird bunch of anti-rebels, born oglers who dare somehow to back away from ironic watching, who have the childish gall actually to endorse and instantiate single-entendre principles. Who treat of plain old untrendy human troubles and emotions in U.S. life with reverence and conviction. Who eschew self-consciousness and hip fatigue. These anti-rebels would be outdated, of course, before they even started. Dead on the page. Too sincere. Clearly repressed. Backward, quaint, naive, anachronistic. Maybe that’ll be the point. Maybe that’s why they'll be the next real rebels. Real rebels, as far as I can see, risk disapproval. The old postmodern insurgents risked the gasp and squeal: shock, disgust, outrage, censorship, accusations of socialism, anarchism, nihilism. Today’s risks are different. The new rebels might be artists willing to risk the yawn, the rolled eyes, the cool smile, the nudged ribs, the parody of gifted ironists, the “Oh how banal”. To risk accusations of sentimentality, melodrama. Of overcredulity. Of softness. Of willingness to be suckered by a world of lurkers and starers who fear gaze and ridicule above imprisonment without law. Who knows.

Perhaps Burnham felt more free to express himself in a more genuine way after having completed his cultural duty as a white man to point out his own privilege before he can talk about anything else. But I think that this “requirement” is absurd. Anyone can and should express their own feelings and experiences in art without having to preface it with some disclaimer about their privilege. It weakens and muddles the work. This is not to downplay the importance of serious social justice issues, but not all things have to address all things. And when serious things are addressed, they should be addressed with seriousness and sincerity, with deliberate strategy, and without ironic deflection.

There is no public square in the modern world. Family and religious structures have been eroding for decades. There is no clear place for anyone to find community or meaning. The most common remaining place for community is the workplace. So it’s perhaps not surprising that The Office remains so popular and relatable to such a wide audience today, over 8 years since its last episode. There are two new podcast series about the show, by the actors who were on the show. Anyone who’s been on dating apps in the past 10 years has seen profiles talking about being “the Pam to your Jim” or vice-versa. And so on and so on.

The Office features a diverse cast of dysfunctional, insane idiots. But the message throughout the entire show is that this is our cast of dysfunctional, insane idiots, who are ultimately still lovable, despite their flaws. Other popular shows like 30 Rock and Parks and Rec follow the exact same format. On the other hand, popular sitcoms of the previous generation — The Simpsons, Family Matters, Full House, Home Improvement, Married With Children — carried a similar message, except with the key difference that this narrative was about actual families rather than corporate workplaces.

Indeed, it is exceedingly common nowadays for a corporate workplace to refer to itself as a “family”. It is only controversial when companies take the opposite stance. Of course, “family” is just a way to trick employees into working harder and longer hours, but nevertheless it provides a welcome sense of meaning for people who have few other sources of meaning in their lives. The Office takes this idea to a comical extreme with its entirely useless group of incompetent paper-pushers wreaking daily chaos. There is no redemption to be had but the sense of familial bond that’s created among coworkers in these “trenches”.

Is the show at some level, therefore, neoliberal propaganda? I am not making that argument. Rather, I am reading the show’s enduring popularity as reflective of people’s yearnings for truer human bonds in a world where those bonds are increasingly hard to find.

We mustn’t be fooled; The Office remains pure fantasy. In reality, even though your boss may be a funny idiot like Michael Scott, they don’t actually hold any underlying warm and fuzzy feelings about you like Michael Scott does for his employees. Real familial feelings are not possible with your boss, because those feelings must be unconditional, but your relationship with your boss and your workplace is fundamentally conditioned upon your productivity. And if we take Venkatesh Rao’s Gervais Principle seriously, then the corporate workplace is a cesspool of sociopaths, and why should we invest our feelings in such a silly and corrupt game?

As for love, the romance of Jim and Pam is also fantasy. In reality, romances in office or professional settings are fraught and dangerous; women can risk career set-backs either way, whether they pursue such romances or not. Meanwhile, a lack of community in general means that yuppies do not have good ways to meet dating partners in less fraught real-life settings. The only option other than dating apps is the one remaining community in which they are involved, the one remaining community that exists for them, the professional community.

I remember that, when I began my professional career, I felt so lonely and disappointed at how I felt so different and alienated from the vast majority of my coworkers. Not that my coworkers were intentionally trying to alienate me, to the contrary, they were always very inclusive and nice. But I just felt that I did not have common interests with most of them; I had low tolerance for the types of lunch discussions that they preferred, talking about where they’d travelled for their vacations, or that new Star Wars movie they’d watched over the weekend; I had little interest in the career aspirations (read: promotions and compensation) that the others seemed so obsessed about.

Contrary to The Office, my solution to this problem has been mostly to compartmentalize my work life and my personal life as separate entities which satisfy different desires and goals. I no longer hope or expect to find deep relationships at work, although I am still grateful on the rare occasions when I do. But I try to view my job just as a place where I make my living and where, if I’m lucky, I also have the opportunity to work on interesting problems that satisfy my intellect, as a bonus. The ultimate meaning in my life is derived elsewhere, in my friends and family, in my personal growth and maturity.

Is it sad to resign to a reality where I don’t expect fulfilling bonds in a work setting? I don’t think it is. I think it is just a fact of life. I don’t think humans found or sought those bonds from their work in prior history, and why should we now?

What is a vibe? It’s a mood, something that you can’t really explain, but that’s captured by a loose association of ideas, concepts, things that you can identify by feeling, rather than by logic. Frank Ocean’s music easily exemplifies the concept, and indeed his debut album Channel Orange even begins with a sample of the start-up music that plays when you boot up the original PlayStation gaming system, surely intending to suggest a nostalgic vibe of lazy days spent gaming. Frank Ocean’s songs have such a strong character of vibe that their remixes have become a sub-genre unto themselves, particularly “slowed and reverbed” remixes that amplify the feelings of ennui in his songs, and that perhaps evoke the effects of drugs and alcohol on perception. The most hyper-specific edits of this sort are those where the reverb is meant to mimic the quality of music playing outside while you’re in the bathroom at a party. A vibe upon a vibe: muted, muffled sounds, perfectly reflecting the repressed longing of Frank’s lyrics.

Vibes are everywhere. Vaporwave and citypop, the vibe of optimistic futurisms of the 80s and 90s. Lo-fi hiphop beats, the vibe of enjoying the warmth of quiet jazzy vintage records in a cozy room with a cat in your lap. Co-Star’s “brutal” push notifications may register on the level of vibes: they concisely capture minute details of users’ realities which make them feel validated or called out — in any case, they feel seen, vibed with. Tinder now has a feature that quizzes you on your vibes. The opening track “DVD Menu” on Phoebe Bridgers’ Punisher employs the same basic trick as the opening of Channel Orange. A man with the username 420doggface went mega-viral for his seemingly narrativeless TikTok of himself drinking cranberry juice while skateboarding and listening to Fleetwood Mac’s “Dreams”.

A recent New Yorker piece on vibes argues that “TikTok provides a rebuke to the truism that people want narratives.” But I don’t agree. The lyrics of “Dreams” speak of a lover who wants their freedom. “Well, who am I to keep you down?”, asks Stevie Nicks. In other words: If you don’t want me, then I don’t want you either, and I’ll enjoy my freedom, too. 420doggface’s TikTok is not just a video of the vibes of a care-free cool guy; it can be interpreted instead as an illustration of a guy who’s decided to not give a fuck about a lover who has scorned him, who’s now enjoying his newfound freedom in his own unique way. There’s a story here, not just a feeling.

How far can we really go with just vibes? I would argue that vibes are a useful aesthetic concept for art, but that they should only be viewed as a basic component of artistic craft. More beautiful, and more true to reality, are layered concepts that are compared and contrasted, if not strictly logically connected then at least narratively connected. To simply do away with such structure would be a terrible loss.

Mozart’s final piece, his Requiem Mass, contains a section titled “Confutatis” which illustrates my point. The lyrics in the original Latin:

Confutatis maledictis
Flammis acribus addictis
Voca me cum benedictis
Oro supplex et acclinis
Cor contritum quasi cinis
Gere curam mei finis

Translated to English:

When the accursed have been confounded
And given over to the bitter flames
Call me with the blessed
I pray in supplication on my knees
My heart contrite as the dust
Safeguard my fate

These lyrics present a vivid contrast between the suffering of the accursed in the bitter flames of hell with the holiness and salvation found in the blessed. These two moods are similarly reflected in the music itself, the former sung menacingly by the deep male voices, the latter sung by the high, angelic female voices. The song does not find its power in a single vibe, but in the structure of two vibes juxtaposed. The pure relief of heaven cannot be fully realized without the stark recognition of the threat of the suffering of hell.

Similarly, the chorus of Elliott Smith’s “Between the Bars” provides a contrast between personal suffering and, in the final line, the comfort provided by a lover, or as the case may be, a bottle of booze.

People you've been before that you
Don't want around anymore
That push and shove and won't bend to your will
I'll keep them still

This contrast again is reflected in the music: both the “people” in the first line and the same “them” in the last line come with an F major chord, and the final word “still” is accompanied by an F minor chord that’s out of key with the rest of the song. This gives a further emphasis to how the last line changes the entire meaning of the phrase and provides a narrative structure.

Peli Grietzer has posited that neural networks can capture the vibe of a work of art. I would agree with this thesis insofar as neural networks can capture statistical patterns and correlations akin to how the PlayStation’s start-up sound matches well with Frank Ocean’s opening lines in “Thinking About You” about his room being a mess, while there is no necessary or logical connection between the two things. Similarly, it is easy to use AI models like GPT-3 to produce artificial writing with great verisimilitude to human writing, at least if you squint.

But when you look closer you often find that there is no real sense that underlies any of it. Although neural networks can in theory capture logical structures underlying concepts, in practice they do not. (Why this happens is a topic for another time.) When provided with the following prompt,

You poured yourself a glass of cranberry juice, but then you absentmindedly poured about a teaspoon of grape juice into it. It looks okay. You try sniffing it, but you have a bad cold, so you can’t smell anything. You are very thirsty. So

GPT-3 completes the paragraph with “you drink it. You are now dead.” The writing style of the passage does suggest a suspicious, suspenseful vibe, like that of a mystery murder novel or a noir film, and perhaps that can explain why the model completes the paragraph in such a grim manner. But the completion is in a logical sense a non-sequitur; there is no natural reason why cranberry juice mixed with grape juice should kill you, nor does the model provide an explanation.

So TikTok’s recommendation algorithm, surely implemented at least in part with a neural network, can probably be accurately described as operating on vibes, a convenient synchronicity with the content itself operating as such. But during those long TikTok browsing sessions, you never feel satisfied. It is a sampling of a vast array of flavors, but without nourishment. You feel the way you used to feel when you would flip channels on cable television late at night. Sometimes the vibe is right, sometimes it isn’t, in the end you always fall asleep exhausted and with nothing to show for it.

“No thoughts, just vibes”, as the memes go. In both culture and statistical learning, the trends ultimately ring hallow to me: resignation to the world as it appears to be, without greater aspiration to its underlying narratives and structures.

Let’s begin with an excerpt from Roger Ebert’s review of the film.

So much happens out of sight in the film, implied but not shown. Noriko smiles but is not happy. Her father passively accepts what he hates is happening. The aunt is complacent, implacable, maddening. She gets her way. It is universally believed, just as in a Jane Austen novel, that a woman of a certain age is in want of a husband. “Late Spring” is a film about two people who desperately do not believe this, and about how they are undone by their tact, their concern for each other, and their need to make others comfortable by seeming to agree with them.

Ebert is correct that much is implied in the film. Regarding the question of “authenticity”, this does read as very Japanese (or East Asian) to me, as there definitely is a culture of indirect, subtle, non-explicit communication, especially around feelings. Even the final twist in the plot (spoiler alert) is based around indirect communication, which from a modern progressive perspective we might even describe as manipulation. This is easily contrasted with the performances in Seven Samurai, where most emotions are clearly displayed or indicated, even in the case of the socially forbidden romance between the young samurai Okamoto and the villager Shino.

But my read on the film’s main takeaways was the exact opposite as Ebert’s. Much like traditional Japanese (and more generally Confucian) culture, the story revolves around the sacrifice of self for fulfillment of duty, whether it is one’s duty in one’s family or one’s duty in one’s society. The film does not necessarily present this as positive or negative; to my eye, it presents it simply as a matter of fact of Japanese society, or of life itself.

Noriko’s character is in fact not progressive, nor is she being beaten into submission by Japanese culture. She seems not so much hindered by traditionalism as an active participant in it. After all, she refers to her father’s widower friend Onodera’s remarriage as “filthy”, and similarly displays great discomfort with the idea of her father Somiya, also a widower, getting remarried. 

Noriko does, however, engage in flirtation with Somiya’s engaged colleague Hattori, spending so much time with him that observers are surprised to find out that Noriko is not actually Hattori’s fiancée. But when Hattori asks Noriko out on a date, she declines despite her true underlying desires, because she does not feel that it would be proper to go out with an engaged man.

Later, Noriko’s eventual agreement to marry Satake happens in parallel with her concession that Onodera’s remarriage is not actually “filthy”. In this light, her decision to marry should be viewed, seemingly paradoxically, not as a conservative decision, but a progressive one. Indeed, perhaps it is the staying in the comfort of the home forever, to never grow up, that would be conservative in this context. To push Noriko out of the nest is, despite the evident ambivalence that she and her father are feeling, the right thing to do in the grand scheme of things.

Ebert again:

“Will you marry?” Noriko asks him. “Um,” he says, with the slightest nod. She asks him three or four different ways. “Um.” Finally, “that woman we saw today?” “Um.” He defends arranged marriages: “Your mother wasn’t happy at first. I found her weeping in the kitchen many times.” Not the best argument for a father trying to convince his daughter to marry.

I also read this scene oppositely to Ebert. The father’s speech was for me the most powerful and profound scene in the film. He speaks of the sacrifice one must make in marriage, that one may be unhappy for one, two, five, even ten years. But:

Happiness isn't something you wait around for. It’s something you create yourself. Happiness lies in the forging of a new life shared together.

One does not have to view this in the bleak way that Ebert does. All relationships have troubles, and while a western notion of romantic love does certainly help, it is not the only facet of a successful marriage. Devotion and the willingness to put the work in can be worth just as much as romantic love. Immediately before Somiya gives this speech, he visibly holds a copy of Nietzsche’s Thus Spake Zarathustra in his hand. Perhaps Ozu intends to reference Nietzsche’s ideas about the overcoming of oneself to create something new — couldn’t that be a marriage?

The final shot of the movie shows Somiya, now having married off his daughter, slumped over in a chair, sadly peeling an apple by himself. We cut to shots of waves at a beach, and the film ends. We have a sense of the Japanese concept of mono no aware, the pathos of the inevitability of loss. Like the waves crash in a predictable, never-ending pattern, so too do the patterns of life itself. The children must eventually leave, and the parents must be alone once again. It is a difficult experience to go through but one must face it as an unavoidable, natural way of life, just as Nietzsche puts it in his concept of eternal recurrence.

My formula for greatness in a human being is amor fati: that one wants nothing to be different, not forward, not backward, not in all eternity. Not merely bear what is necessary, still less conceal it—all idealism is mendacity in the face of what is necessary—but love it.

Ozu does not provide any hint whatsoever as to whether Noriko will be happy in her marriage. But in the end, no one can really know or predict whether they will be happy in the long term, whether it is in a love marriage, an arranged marriage, or a life of being single. Either way, one must face the future with equal parts courage, joy, and acceptance.

In 2018, the BBC polled film critics on the greatest “foreign language” films of all time. By “foreign language” here they meant non-English, but the critics were selected internationally, so I don’t know why they didn’t just say non-English, but okay, let’s let that slide for now.

The top-voted film in the poll was Akira Kurosawa’s Seven Samurai (1954). Interestingly, though, exactly zero of the Japanese film critics voted for it. Instead, among Japanese films, they preferred the works of Ozu and Mizoguchi. It’s a weird situation: Seven Samurai is the uniformly critical consensus greatest film of all time abroad, and yet in his own home country Kurosawa’s work doesn’t even appear in critics’ top ten lists.

Or maybe it’s not so strange: Different cultures just like different things for different reasons. Apparently, there’s been a whole long-running debate in Japan over whether Kurosawa’s films are “too western”, whether they’re “authentically Japanese”. This sort of thing happens in the opposite direction, too, though maybe not to such an extreme extent. For example, for bands that don’t make it in the west, there’s always the possibility of making it “big in Japan”.

There certainly exist gulfs between cultures. How can they be crossed? I am undoubtedly a western person for most intents and purposes. But I don’t want to have a totally western view.  I’d hoped that perhaps watching Seven Samurai could be a step in the direction towards understanding something about Japan and the Japanese perspective, but funnily enough, it might not be.

Like Kurosawa, I make mad films, okay, I don’t make films
But if I did they’d have a samurai
Gotta get in tune with Sailor Moon
’Cause that cartoon has got the boom anime babes
That make me think the wrong thing

How sad would it be to have a conception of a foreign culture as simplistic as those in the Barenaked Ladies’ lyrics? I want to be able to understand different cultures and their tastes at a deeper level. In part this is for political reasons, but it’s also for personal and aesthetic reasons. On a personal level, I want to be able to better understand the China from whence I myself was born. On an aesthetic level, it seems analogous to not being able to see certain colors, or not tasting certain flavors, or not having certain thoughts; why should I limit my experiences in life?

There isn’t anything inherently bad about the west — I’ll even admit that I like the Barenaked Ladies song. Kurosawa’s films being “western” — if it is even correct to characterize them as such! — doesn’t mean they’re bad, just that they happen to appeal to westerners. “Authenticity” does not inherently signify goodness, nor does “appropriation” inherently signify badness. The different tastes can simply co-exist and do not have to be in competition with one another, despite the tacit assumption of singular objectivity underlying a ranking of the “greatest” films.

But it is important to not conflate the appropriated remixes viewed through a western gaze with the originals viewed through native eyes. They are two different things. Neither one more “authentic” than the other, they just are what they are in their different contexts. 

As Noah Smith recently wrote about weebs (a bizarre 4chan neologism that means “western Japanophile”):

Japanese people themselves tend to be somewhat mystified when they encounter weebs. It’s a very odd thing to have your country’s pop culture used as a platform for a bunch of foreigners to create a real-world facsimile of a romantic youth they imagine you got to enjoy. And perhaps this is why weebs don’t often move to Japan (and those who do often find it unsatisfying); for most, it’s not the place they really want to go.

Because Japan isn’t actually the place that lets weebs — or at least, the ones who aren’t too shy to leave their rooms — live the romantic dream. That place is weeb culture itself. America and other countries where the culture has taken root didn’t ape Japan; they created something entirely original.

It is certainly easier for us Americans to access the spin-offs, the fortune cookies, the California rolls, the hard-shelled tacos, if only simply because they’re more available. But the originals remain harder to grasp. They cannot just be consumed, we must also engage with them. We know our own culture to be a living, dynamic, evolving thing, and of course the same is true for other cultures.

Even as observers, an obvious initial barrier is language. I am not saying, a la Sapir-Whorf, that this is a necessary barrier, or that there are untranslatable thoughts. The issue is more that, a la “Manufactured Consent”, publishers are incentivized to put out the material that will sell, in this case, stuff that already happens to be more aligned with a western sensibility. So, for simple economic reasons, we’re more likely to see the Kurosawas, and less likely to see the Ozus and the Mizoguchis. If we seek more, then there’s no way around it, we must commit to the work of learning the foreign language or culture.

In his poem “Do You Speak Persian?”, Kaveh Akbar writes about the sadness of losing connections with his family and past through the loss of language.

I have been so careless with the words I already have.

I don’t remember how to say home
in my first language, or lonely, or light.

I remember only
delam barat tang shodeh, I miss you,

and shab bekheir, goodnight.

How is school going, Kaveh-joon?
Delam barat tang shodeh.

Are you still drinking?
Shab bekheir.

I don’t want to be so careless. The cost of learning a foreign language can be high, but I can’t imagine that it’s not worth it, as it opens access to an entire parallel universe. Some may be captivated by Elon Musk’s mission to Mars, but I prefer to visit the other worlds that can be found right here on our home planet. The BBC so easily and casually falls into an Anglocentric framing in its films poll, but we should aim to make the “foreign” less foreign.

The photos are just as my parents had described it to me: There was hardly any space, and even if you did get a seat it would be hard and uncomfortable. This man seems to have discovered a way to rest his feet, despite not having a seat. You take what comforts you can.

You’d be even luckier to have a bed. Sleeping on the floor was common. Here is a family relaxing together on the floor directly next to the bathroom. One child sleeps peacefully against the bathroom door, and the other child is contentedly resting in his father’s arms. Meanwhile, a man smokes a cigarette in the foreground — the commotion and the crowd is always there.

I remember visiting China as a child in the 90s and there being no streetlights near my grandfather’s house. I remember riding on the back rack of my uncle’s bike in the absolute pitch black of the night. On long distance trains, my family would put me in the comfortable and spacious first class cabins, thinking of me as a fragile American who couldn’t handle the cramped, loud, chaotic quarters of the rest of the train. Naturally, I’ve always romanticized that forbidden experience portrayed in Wang’s photography. 热闹, rè nao, hot and noisy — perhaps uncomfortable, but full of excitement and life.

In 2018, I flew into Shanghai Pudong airport and took the train to Suzhou. My grandmother, still imagining that the train would be like the trains in these photos, was terrified that I’d get robbed or scammed. But they were nothing like that. They’re super clean, super modern, 180mph bullet trains, far nicer than anything in America, which has at this point seemingly no hope of developing anything remotely comparable in terms of public infrastructure.

Visiting the supermarket was just as impressive. The selection of produce exceeded the selection in any US supermarket tenfold. There were also more exotic, absurd flavors of potato chips than you could possibly imagine. Cream soda flavored chips? Cucumber flavored chips? Numbing spicy hot pot flavored chips? Yeah, you got it.

I wonder how my parents felt when they first came to America in the 80s, I wonder if they felt impressed by the highways and streetlights and wide supermarket aisles. Saving up to buy that house in the suburbs, with all the space they could ever want, but none of the bustling energy of home, for better and for worse. The quiet with which my father could do his work without the constant disturbance and uncertainty of social unrest. The loneliness and isolation that crushed my mother. I think about how I didn’t make it any easier, I think about the grief that I gave them when I looked down on them as a teenager for not understanding me, my culture, the American point of view that I so arrogantly knew to be superior.

We imagine the United States as the shining promised land of the world. Immigrants are supposed to just be thankful to have the freedom and stability that this country claims to provide. It’s never so simple. While we long for a more beautiful future, we also long for the beautiful past.

I would like to ask my parents: Was it worth it? But I know it’s not the right question to ask. You can never really go back. Even if you do go back, it won’t be the same. Nothing stays the same. You can only ever move forwards, no matter where you are.

All short-term studies have shown the COVID vaccine to be safe and effective. It is impossible at this time to know about the long-term effects, but the situation is being monitored very closely.

It is certainly true that we do not yet have any data about the long-term effects of the vaccine, and we can’t have that data until a longer period of time has passed. But does it follow that it’s impossible to know anything about the long-term effects until then?

Ways of knowing

It is a classical question in philosophy: How do we know things? The two main sides: rationality, which argues that human reason itself can determine truths about the world, and empiricism, which argues that evidence gathered through our senses is how we acquire knowledge about the world.

Both sides are true. But in today’s age, especially in view of “big data”, the empiricist view is increasingly prominent. It has become standard practice to ask for empirical evidence to justify claims. Randomized controlled trials (RCTs) are the gold standard of rigor everywhere, from economics to social science, to medicine, to feature development for apps and websites. The word science has become a near synonym for the practice of RCTs.

And yet RCTs don’t tell you everything. A famous example is Thalidomide, which was used by pregnant women in the 50s and 60s to treat morning sickness, but it had the tragic side effect of babies born with deformed limbs and organs. To explain this outcome, a common narrative you’ll hear is that the drug had never been tested on pregnant women.

Digging deeper into the history, it turns out that not testing on pregnant women was not the only problem. The drug was synthesized by a West German pharmaceutical company in 1954 searching for a cheaper method to manufacture antibiotics. They created Thalidomide only incidentally, and began to “cast about for a disease that thalidomide could cure”. The drug was found to be useful as a sedative, and was shown to be safe in animal trials. But despite it having been to some degree empirically tested, the drug was still not in any way understood.

How did it operate? No one knew.

Why vs. whether

Science is not only the practice of doing experiments. Empiricism can be used to confirm whether something is true or false. But it cannot be used to explain why something happens or does not happen.

A clear illustration of this dichotomy can be found in machine learning research, where often the objectives are stated purely statistically. But adversarial examples can be found that trick models into catastrophically misclassifying stop signs as 45 mph speed limit signs. Ultimately, the reason is because the models are only empirically evaluated on a predefined dataset, and not in any holistic way that tests for the understanding that a stop sign is an octagonal red shape with white text that reads “STOP”. As Noam Chomsky stated in his critique of statistical learning,

[This] is a notion of success which I think is novel in the history of science. It interprets success as approximating unanalyzed data.

The same critique has been made by Nobel laureate Angus Deaton in the context of economics and foreign aid.

Thirty years of project evaluation in sociology, education, and criminology has been largely unsuccessful because it focused on whether projects worked instead of on why they worked.

Experiments tell you whether an intervention works in a given context, but not necessarily more than that. If we consider an economic intervention as a black box, then if we show that it works in one country, we cannot necessarily conclude anything about its efficacy in another country, for we may have no grounds to equate the two, given their different histories or other complexities.

Similarly, if we treat a drug as a black box, when it is demonstrated to work in one situation (e.g. on rodents), then it cannot necessarily be concluded that it is effective in another (e.g. for pregnant women).

We mustn’t rely solely on empiricism; we need rationality and reasoning as an underlying structure, as a stabilizing narrative. Understanding is what gives us confidence in the robustness and generalizability of our knowledge.

The vaccine

The COVID vaccine is decidedly not a black box. It is a feat of engineering; how it works is well-understood. However, the health authorities have not done a good job of communicating this understanding to the public, and often instead argue via appeal to authority — Believe Science!

Information on how the vaccine actually works exists and is available. Artificially sequenced mRNA is injected into your cells; it encodes a sequence of amino acids for your ribosomes to follow to produce an approximation of the COVID spike protein; the immune response then learns to recognize that protein, and so also the actual COVID spike protein.

So will the vaccine have adverse long-term effects? With a better understanding of the vaccine, we now have the ability to engage this question more pointedly. Each component can be taken apart and scrutinized. Could the mRNA itself have any long-term adverse effects? How long does it stick around in your body? What about the approximated COVID spike protein? How do we know that the immune response will be able to generalize to the actual COVID spike protein? Does the mRNA produce any other byproducts? I don’t have answers to these questions, but you can hopefully see that these questions are answerable, and their answers can provide us with more (or less) confidence about the vaccine, without necessarily requiring any new empirical tests.

Officials may be wary of providing such information because they feel that it can only provide ammunition for conspiracy theories. But perhaps it is instead the withholding of such information that fosters conspiracy theories. And for something as important as the COVID vaccine, we should be encouraging rather than discouraging public discourse and understanding. The story is more complex than just believing in some numbers or statistics. It always is.