Recently on Twitter, Jeffery Hinton summed up an ongoing dispute he's had with Yann LeCun as follows:
The central issue on which we disagree is whether LLMs actually understand what they are saying. You think they definitely don't and I think they probably do.
In the thread that followed, several commentors asked what, exactly was the notion of “understanding” that Hinton was appealing to here. What, exactly, are we saying when we say that an LLM “understands what it is saying”? In current debates about whether LLMs reason or understand, there is very little clarity with respect to this question. In this post, I want to articulate a conception of what it is for something to “understand what it’s saying” that I hope will bring clarity to this debate. The answer at which I'll arrive is that current LLMs at least sort of understand what they're saying sometimes, and future LLMs, even those that are the product of simply scaling current methods, could fully understand what they're saying. However, the main upshot of this post is not meant to be an answer to the question itself, but, rather, the general approach to answering it and similar questions.
Let us start with some technical preliminaries. LLMs are neural networks trained to predict the next word in a sequence. An LLM works in tokens, numerical encodings of the semantically relevant bits of words. Though tokens cut more finely than words, for simplicity, we can just treat tokens as numerical encodings of words. What it gets, in its training data, are strings of tokens from all across the internet, and its task, in training, is to learn how to complete these various strings. For instance, it might get the string of tokens, “3742 1569 8320 4915 2876 ____” with the last token missing from the string. Through its training run, it learns to assign a probability distribution to the various tokens that might complete this string, eventually learning to predict that the token most likely to come next is 6498. Completed with that last token, that string might encode the sentence “The cat sat on the mat.” This ability to predict the next word translates to the ability to speak a language since, if you can predict what human beings would say, you can simply start speaking, predict the next word that comes out of your mouth, say that, and iterate this process. Doing this, you'll end up speaking in sentences that humans would actually say, making good sense to other humans. It is an amazing fact, that through training on nothing but next-token prediction, an LLM comes to be able to do this, at least seeming to acquire the ability to speak a language, or "linguistic competence."
There are several aspects to the sort of “linguistic competence” that an LLM comes to acquire. It is able to tell, for instance, that “3742 1569 8320 4915 2876 6498” is a grammatical string, something that can be meaningfully used, as is “3742 6498 8320 4915 2876 1569,” but “2876, 4915, 1569, 8320, 3742, 6498” is not. The ability to sort strings in this way is a matter of syntactic knowledge, classifying the different tokens into different syntactic types and knowing the rules by which different syntactic types can compose to constitute meaningful bits of language. It is more or less uncontroversial that LLMs have some sort of syntactic knowledge. The much more controversial question is whether they have semantic knowledge: not just knowledge of grammar, but knowledge of meaning. Whereas syntactic knowledge involves knowing, for instance, that the tokens 1569 and 6498 are both of the type COMMON-NOUN, semantic knowledge would involve knowing, more determinately, that 1569 means cat and 6498 means mat. When we speak of a system as "Understanding what it's saying" we are principally attributing to it this sort of semantic knowledge. The question is whether we can really make such an attribution. How we answer this question will depend on how we think about semantic knowledge and what we’re doing when we attribute it.
Standard approaches to semantic knowledge think that what one is doing in saying of something that it “understands what its saying” is ascribing to it some specific kind of representational state: very roughly, a state of associating the symbols of the language it's using with things in the world of which one has some representation, for instance, associating the token 1569 with cats. To answer the question of whether an LLM “understands what its saying,” is to find out whether it instantiates this specific sort of representational state when it produces sentences such as “The cat is on the mat.” On this standard approach, the question of whether something instantiates such a state is essentially an empirical question. Though we might infer that something instantiates this state on the basis of how it behaves, the state we are actually reporting when we say that something understands what it's saying is something that we'd have to look "under the hood" to find, whether in the brain, in the case of human beings, or in the execution of the program, in the case of an LLM. Now, given how fundamentally different the working of a human brain is from the working of an LLM, if one has this general approach, one is likely to come to the conclusion that LLMs don't instantiate anything like the internal state humans instantiate when then understand what they're saying. Accordingly, LLMs don't really understand what they're saying. I want to suggest here, however, that this approach to thinking about semantic understanding is in fact radically mistaken. Let me explain.
To Place or Not to Place In the Space of Reasons
The core point I want to make here is that, in saying of something that it "understands what it's saying," one is not giving an empirical description of this thing. I draw my inspiration in making this claim from Wilfrid Sellars, who famously said that the question of whether someone knows something is not an empirical question. In his master work, Empiricism and the Philosophy of Mind, Sellars famously wrote:
In characterizing an episode or state as that of knowing, we are not giving an empirical description of that episode or state; we are placing it in the logical space of reasons, of justifying and being able to justify what one says.
Sellars's basic idea here, elaborated at length in the work of Robert Brandom is that, in saying that someone knows something, you’re taking them to have a certain sort of authority in their making of a particular claim. You take them to be entitled to that claim, and able to bear the justificatory responsibility of demonstrating this entitlement in response to appropriate challenges. Accordingly, you take it that you can make this claim yourself on the basis of their authority, able to defer back to them in response to a challenge. Such a “taking” is not the taking of an empirical fact to obtain. It is, rather, a normative taking; the adoption of a complex normative attitude with respect to someone's making of a claim.
Now, Sellars is talking primarily about empirical knowledge here—knowledge of how things in the world are. However, the basic point applies just as well to semantic knowledge—knowledge of what one says in uttering some sentence. Whereas attributing empirical knowledge to someone is taking someone to be able to respond for demands for empirical reasons, actually providing reasons to justify what they've said in response to appropriate challenges, attributing semantic knowledge to someone is taking them to be appropriately responsive to reason relations as such, appreciating, for instance, what counts as a challenge to what one has said (whether or not one can actually respond to that challenge). Consider the case of someone who says "The ball is red." Someone who says such a thing commits themself to the claim that it's colored, precludes oneself from being entitled to the claim that it's white or gray, and so on. These are the reason relations one binds oneself by in saying that the ball is red. Unless one recognizes that one has bound oneself by these reason relations, one does not understand what one has said. Attributing this understanding to someone, then, is not taking a particular state to obtain in their brain, but, rather, taking them taking them to be responsive to demands for reasons. Let me illustrate this idea with two examples.
Suppose I go to China. I don’t speak any Mandarin, and someone tells me to say “qiú shì hóngsè de.” If I manage to say this (somehow getting the pronunciation right), I will be regarded as committed to “qiú shì yǒu yánsè de,” precluded from being entitled to saying “qiú shì huīsè de,” and so on. However, I won’t have any knowledge of the fact that I’ve bound myself by these reason relations. Accordingly, I won’t be responsive to demands for reasons in Mandarin. Someone who speaks to me, questioning what I say, will quickly realize that I simply can’t be held accountable for anything I say. In that sense, I don’t know what I’m saying. The important point here is that to say this is not to make an empirical description of me. It’s not to say that there’s anything going on or failing to go on in my brain. It is, rather, a normative thing: recognizing that I’m not responsive to reasons asked for or given in Mandarin. My utterances aren't to be counted as moves that I am making in a Mandarin-speaking discursive practice.
Here's a different kind of example. Suppose a parrot, having heard people speak, squawks out “Brawk, red ball!” Does the parrot understand what it's saying in squawking out such a thing? Clearly not. Once again, the key thought here is that, in saying this, we are not ascribing any specific properties to the parrot's brain---saying of it that it lacks some sort of representational state. Though, of course, there are various facts about the parrot's that can be appealed to in order to explain why it behaves as it does, what we're doing in saying that it "doesn't understand what it's saying" is not describing any such facts. Rather, we are refusing to situate the parrot's squawk in the space of reasons, thinking of it as a "move" in the "game of giving and asking for reasons." We don't count the parrot as bearing any justificatory responsibility for its various squawks and squeals. That's why its various squawks and squeals do not actually amount to its saying anything at all.
Now, there’s an important difference between the case in which I say something I don’t understand in a language I don’t speak such as Mandarin and a case in which a parrot squawks out a sentence, where it shouldn't really even be counted as speaking at all. Suppose someone tells me to say something very offensive in Mandarin, and I say it, not knowing what it is that I’m saying. I can still be held accountable for what I’ve said, even though I don’t know what it is. That’s because, though I am not responsive to reasons in Mandarin, I am still responsive to reasons in general, and so I’m still a bearer of justificatory responsibility generally. Accordingly, we can ask such things of me as “Why would you that, if you don’t know what you’re saying?” We can’t, of course, ask any such thing of a parrot. Thus, while it’s reasonable to regard me as saying something, but just not understanding what I’m saying, the parrot is not regarded as saying anything at all. Still, the general point about the normativity of these claims holds of both cases.
The Case of LLMs
Let us now turn back to the main topic of this post: LLMs. On this approach, to ask whether an LLM understands what it’s saying is to ask whether, unlike a parrot, we can hold it responsible for what it says, counting on it to give reasons, to respond to potential challenges, and so on. Clearly, on this front, it fares much better than a parrot who merely squawks out sentences. When ChatGPT says something, you can ask follow-up questions, and it generally responds appropriately: clarifying, qualifying, responding to potential challenges, and so on. Consider this exchange, in which I ask it about the colors of balls, or this exchange in which I ask it about cats being on mats. Here, it at least seems to exhibit an understanding of what it's saying when it says “The ball is red.” Does it really understand what it's saying? What are we asking here? The point I want to emphasize is that we are not asking about whether some process is going on “under the hood.” Rather, we are asking whether it is really appropriately responsive to the reason relations that it binds itself by in saying "The ball is red." Insofar as this is our question, given the sort of reason responsiveness it exhibits, think we can reasonably attribute to it at least some level of semantic knowledge. That is, I think we can say that it at least partly “understand what it’s saying" when it says, for instance, that something is red.
Why the qualification of "at least partly"? The reason is that things are not always so clear-cut. To see a case in which GPT4 fails, consider the following question, which I've gotten from the YouTuber Mathew Berman:
A small marble is put into a normal cup and the cup is placed upside down on a table. Someone then takes the cup and puts it inside the microwave. Where is the marble now? Explain your reasoning.
An ordinary speaker, who knows that normal cups are closed at the bottom and open at the top will judge that, if just the upside-down cup is picked up, the ball will stay on the table. I take it that our drawing this inference is a matter of semantic knowledge, our understanding of the meaning of "cup" and "ball." GPT4, however, does not seem to possess this understanding. In this exchange it suggests that its default understanding of "cup" is something that is closed at both the top and the bottom, like jar. However, when, in a new context, I ask the question again, clarifying that a normal cup is open at the bottom and closed at the top, it still gets the wrong answer, maintaining that, somehow, the marble remains in the cup due to gravity. On the basis of these failures of reasoning, I think there's reason to think that it doesn't fully understand what its saying when it says such things as "The ball is in the cup."
So, to the question of whether current LLMs really "understand what they're saying," it seems clear that the answer is (perhaps unsurprisingly): sort of . . . sometimes. However, while this question can't be answered simply in the affirmative or negative for current models, I hope I've done enough to show how we should approach this question, when thinking about future models. Whether or not we should treat a system as "understanding what it's saying" is not a matter of whether there is anything like the the processes going on inside of it that go inside our brains when we speak and understand what we're saying. It is, rather, a matter of whether it manifests an understanding of reason relations it binds itself by in saying what it does in responding to queries and challenges. Though, as the case of the ball in the cup illustrates, current models do not always manifest such an understanding, I do not see any reason in principle why future models, even those that are the result of nothing more than the scaling of the techniques of current models, could not. If they do, we would have every reason to say that those LLMs really do understand what they're saying.
What’s the meaning of the word “red”? There are different approaches one might take to answering this question.
A representationalist approach to linguistic meaning will answer this question by saying that “red” represents a specific non-linguistic quality, namely, redness. But what is redness? One might think that no answer to this question can be given. Clearly, however, we can at least provide a partial answer. We can say, for instance, that redness is a color, that something’s being scarlet or crimson implies that it is red, that something’s being red (all over) is incompatible with its being green (all over), and so on. Saying such things, one specifies, in relational terms, what it is for something to be red. The core idea of an alternative approach to linguistic meaning, known as inferentialism, is that, in saying such things, all one is really doing is expressing the inferential rules governing the use of the term “red,” and it is in terms of these rules that the meaning of “red” is to be understood.
An inferentialist approach to linguistic meaning understands the meaning of a word in terms of how it contributes to the meanings of sentences, understanding the meanings of sentences in terms of how they inferentially relate to other sentences, most fundamentally, their implying and being implied by other sentences and their being incompatible with other sentences. For instance, “x is red” implies “x is colored,” is implied by “x is scarlet” or “x is crimson” is incompatible with “x is green,” and so on. Clearly, however, such a purely inferential specification of the meaning of “red” can’t suffice to completely specify its meaning, right? Almost all so-called “inferentialists” will answer this question by saying “Of course, not!”
According to the standard version of inferentialism, which I’ll call “quasi-inferentialism,” accounting for the meaning of a sentence requires appealing to more than just inferential relations between sentences. On the quasi-inferentialist account, there must be, in addition to inferential relations between sentences, relations between perceptual states and sentences—so-called “language-entries”—as well as relations between sentences and intentional actions—so called “language-exits.” Focusing just on the case of perception, it’s clearly essential to the meaning of “red” that one can come to know that something’s red, thus being in a position to correctly apply the word “red” to that thing, by seeing that it’s red. For instance, one can come to be entitled to the claim “The ball is red” in virtue of seeing the following red ball:
The quasi-inferentialist accounts for this aspect of the meaning of “red” by including in their semantic theory the following “language-entry rule”:
Quasi-inferentialism, however, faces a basic problem. It is advertised as an account of the meaning of words like “red.” However, in spelling out the account, we end up using the word “red” to specify the circumstance under which one is entitled to use the word “red.” We’re thus appealing to the very meaning for which we’re supposed to be inferentially accounting in giving our account. Thus, if we really try to give an account of the meaning of “red” along these lines, our account would be circular.
Of course, one might think that there’s no non-circular account of the meaning of words like “red” to be given, and so we should simply accept that we’re always going to appeal to the meaning of “red” in articulating our account of its meaning. If one thinks that, however, there is little reason to adopt any sort of substantive theory of meaning at all, be it inferentialist or quasi-inferentialist. Rather than trying to articulate the meaning of “red” in terms of the complex set of rules governing its use, one might as well just state one very simple rule that perfectly suffices to capture it’s use:
To resolve the problem, let us return to the starting thought that it’s essential to the meaning of the word “red” that one know that something’s red by seeing it. My basic positive proposal is simple. Rather than attempting to account for this aspect of the meaning of the word “red” by appealing to a quasi-inferential relation between a perceptual circumstance and the use of the word “red,” we can account for it in purely inferential terms simply by articulating the inferential relations between “red,” “sees,” and related terms. Let me explain.
In articulating what it is for something to be red, we might say such things as that if one is in a position to see that some object x is red—looking at x in good lighting—and one has color vision, then one will see and thereby know that x is red. The core inferentialist thought, applied in this case, is that we can think of what we say here as the expression of inferential rules governing the use of “red.” Spelling this out explicitly, we can inferentially define the predicate “is positioned to see that x is red” in terms of inferences like the following:
I’m aware that this account of the meaning of “red,” which appeals to nothing but inferential relations between sentences, is going to be met with skepticism from most readers. Let me address some concerns and articulate some consequences.
One initial worry is that, on this account, there will be no way to distinguish the meanings of words like “red” and “green.” One might wonder about a mapping of the language onto itself where “red” is and “green” are switched with all of the inferential relations being preserved, for instance, with “crimson” being mapped to “forest green,” “colored” being mapped to itself, and so on. Of course, it’s true that, for all of the inferences I’ve specified as examples thus far, this is possible. Yet, on this account, the meanings of these words are understood in terms of the entire web of inferences, which includes more than just relations between color words, and, in the context of this whole web, they are indeed distinct. For instance, “x is a tomato” along with “x is red” implies “x is ripe” whereas “x is a tomato” along with “x is green” implies “x is unripe.” By inferentially connecting color words such as “red” and “green” to non-color words like “ripe” and “unripe,” we can account for their distinctive conceptual significance. I acknowledge that one is likely to feel that something must be left out of this account. I had this feeling too when I first started developing this theory. My claim, however, is that nothing is left out.
To make this claim vivid, consider Mary, the color scientist who’s been in a black and white room since birth and so has never experienced color, but has reached the theoretical limit of what can be known about the colors without actually having experienced them. Most people have the intuition that she does not know what it is for something to be red. I disagree. On this account, she knows just what it is for something to be red, since she grasps all of the inferential relations between sentences that articulate the content of “red.” Though she’s never herself used the term non-inferentially, she knows just the conditions under which it can be non-inferentially used, and that is all that is required in order to be counted as knowing the meaning of “red” on this account. Of course, one might be inclined to just pound the table and assert here “But she doesn’t know what it is for something to be red!” However, I don’t see a non-question-begging argument for this claim.
Here’s another consequence of this view of meaning. There has recently been much debate about whether a Large Language Model, trained on nothing but linguistic data, without any sort of “sensory grounding,” could actually understand natural language. Some philosophers and computer scientists have argued that such a model can’t understand language at all, whereas others have argued that such a model could only understand a specific subset of natural language expressions, for instance, those belonging to pure mathematics and logic. This account has the radical consequence that a model trained on nothing but linguistic data could in principle grasp—completely grasp—the meanings of all natural language expressions, even including those that are essentially such as to be deployed perceptually such as “red.” Once again, on this account, the meaning of “red” is constituted—completely constituted—by the inferential relations sentences containing it bear to other sentences, and a Large Language Model is in principle capable of grasping all such relations. Understanding how such an expression can be non-inferentially deployed is an essential aspect of grasping its meanings, but actually deploying such an expression non-inferentially is not. Once again, this consequence of the account might be unintuitive to many, but I don’t see a non-question-begging argument against it.
I develop this argument for this radical form of inferentialism, which has been termed “hyper-inferentialism,” in my recent paper How to Be a Hyper-Inferentialist. Check it out, if you're interested in hearing more of the details!
One of the shows that has provided the most philosophical food for thought over the years is Rick and Morty. Love it or hate it, you can’t deny its creativity. The show has become more and more mind-bending throughout the seasons, but, in the very first season, we are introduced to one of its most curious characters: Mr. Meeseeks. It's immediately apparent that Mr. Meeseeks is very unlike ourselves. But I want to suggest here that he is even more radically unlike ourselves than we might initially think. Ultimately, I want to suggest that getting into view what it is to be a Meeseeks, where being a Meeseeks is radically unlike being one of us, might help us get clearer on what we ourselves are.
The Curious Case of Mr. Meeseeks
One of the Rick’s most memorable contraptions is his “Meeseeks Box.” It’s a cube with a big button on it. When you press the button, a blue individual who calls himself “Mr. Meeseeks” pops into existence and asks what he can do for you. You give him a task, for instance, opening a mayonnaise jar, and, once he does it, he ceases to be, popping out of existence. This, if all goes well, is how the life of a Meeseeks goes. About their lives going this way, Rick says “Trust me--they’re fine with it.” Assuming Rick is telling the truth here, how could this possibly be? How could one possibly be fine with one’s life going this way, popping into existence only to complete some task and popping out of existence as soon as it’s completed?
The Rick and Morty fan wiki says a Meeseeks only does what a Meeseeks does because a Meeseeks is in great pain. Some of the lines in the show suggest this way of thinking. Towards the end of episode, one of the Meeseeks says “Existence is pain to a Meeseeks, and we will go at any length to alleviate this pain.” This way of thinking makes quite a bit of sense to us. We can imagine a life in which we’re born into tremendous pain, so much pain that we’d rather die than continue to live in such pain, and we’d go to great ends to alleviate the pain. Still, this just doesn’t seem like an apt characterization what it is to be a Mr. Meeseeks.
Meeseeks are characteristically cheery. Why would Mr. Meeseeks be so cheery, at least initially, if he was in tremendous pain, and his only motivation was to alleviate this pain? You would think that life would just be agonizing for a Meeseeks from the outset. But that seems to not be so. They come into being and they seem happy to be here. They are given a task and they seem to love to do it, to get it done, and to cease to be. Though the particular Meeseeks who are burdened with the seemingly impossible task of taking two strokes off Jerry’s golf game express being in a state of agony, this is an uncharacteristic state for the Meeseeks to be in. Meeseeks are generally cheery, not in miserable agony. Of course, they could be pretending to be cheery when, in fact, they’re in miserable agony, but we don’t have any reason to think that that’s the case. We need an alternate way of making sense of what it is to be a Meeseeks. For that, I suggest we turn to Aristotle.
Some Aristotelian Ontology
Our question is: what is it to be a Meeseeks? What kind of science can we look to for an answer to a question like this? If our question was “What is it to be an orca whale?” or “What is it to be a queen hornet,” then we could look to cetology, the branch of zoology that studies whales, dolphins, and porpoises or entomology, the branch zoology that studies insects. But there is no branch of zoology that is apt to answer the question of what it is to be a Meeseeks. Zoology deals with creatures of Earth, and, wherever Mr. Meeseeks is from, it’s certainly not Earth. Even the more general science of biology will not be of help, since, once again, biology still deals with living organisms that belong here on Earth, and, once again, Mr. Meeseeks is not from here. If there is a science that will help us understand what it is to be a Meeseeks, it must not be tied to any possibly contingent way that things are here on Earth. It must consider what it is for something to be, in the myriad ways in which being can be done. The activity of being, for Mr. Meeseeks, is quite unlike the activity of being for any of us living beings here on Earth. However, it is not unintelligible; we can make sense of it, but, to do so, we have to draw on the most basic science that there is: the science of being qua being, which studies what it is for something to be, not insofar as it a cetacean, an insect, an animal, or even a living organism, but, rather simply insofar as it is---insofar as being is something that it does.
The science of being qua being is not a science they teach in grade school anymore. But I’d like to think that it really is a science, in the proper sense of the term. One person who certain did think this was the greatest of the Greek philosophers: Arsitotle. Aristotle was a true polymath. He had many works in different sciences, mathematics, biology, astronomy, among others. But one science had a special place in Aristotle’s heart: the science of being qua being, the topic of perhaps his greatest work, The Metaphysics.
I’ve said that being is an activity. This is one of the fundamental claims of Aristotle’s Metaphysics. An activity, in the broadest sense of the term, is something that one does. But not all things that one does are of the same kind. In Book Theta of the Metaphysics, Aristotle makes a distinction between two fundamentally distinct kinds of activities, that I think will enable us to get clear on what it is to be a Meeseeks. The Greek words that he uses for these two kinds of activities are “kineses” are “energeiai.” I could provide a translation of these words, but all translations are controversial, and picking any one of them, I think, would do more harm than good in understanding what they mean. They are best understood directly by example.
When Rick first presents the Smith family with the Meeseeks box, he pushes the button and tells the newly existent Meeseeks to open Jerry’s stupid mayonnaise jar. Consider this activity, the activity of opening Jerry’s stupid mayonnaise jar. This activity has what Aristotle would call a “telos,” an aim. The aim of the activity of opening the jar is to have opened it. Once you’ve opened the jar, you’ve done what you’ve aimed to do in opening the jar; you’ve gotten it open. The activity of opening the mayonnaise jar is an activity such that, once the aim is achieved, the activity is no more. Once you’ve opened the mayonnaise jar, you’re no longer opening it. You've done it, and so you’re no longer doing it. This is the logic of an activity of the sort that Aristotle calls a “kinesis.”
A kinesis is an activity with an aim such that, once the aim is achieved, the activity ceases to be. In this way, a kinesis might be thought as an activity that aspires to its own non-being. For the activity to be successful is for it to have achieved its aim, and, once it does that, it is no more. So, the aim of a kinesis is the termination of the kinesis. Some other clear examples of kineses are fixing the dishwasher, doing your math homework, and climbing down the courtroom steps. The fact that each of these activities is a kinesis can be shown by the following test: the truth of a statement that uses the progressive tense of a kinesis-expressing verb entails the falsity of the statement that uses the perfect tense of that verb. For instance, insofar as you’re fixing the dishwasher, you have not yet fixed it. Insofar as you’re still doing your math homework, you have not done it. Insofar as you’re still climbing down the courtroom steps, you have not yet climbed down them. All of these activities have, as aims, the getting done of the doings that they are. So, they all have, as aims, their own non-being.
Now consider the activity of dancing. Now, in some cases, you might dance with the aim of winning a dance competition, or with the aim of impressing a dance partner, but, most of the time, when we dance, we do it just to do it. Usually, we just dance to dance, with no aim other than the dancing itself. When we think of dancing in this way, we see that it is an activity quite unlike the activity of opening the mayonnaise jar. Recall, insofar as you’re opening the mayonnaise jar, you haven’t yet opened it, and, once you’ve opened it, you’re no longer opening it. That’s not so with dancing. Insofar as you’re dancing, you’ve danced, and just because you’ve danced, it doesn’t mean you’re no longer dancing. Unlike opening the jar, dancing is an activity whose end is internal to itself. There’s nothing that one aims to accomplish in dancing to dancing that is external to the activity of dancing itself. Thinking of dancing in this way, as an end in itself, dancing is a sort of activity that Aristotle calls an “energeia.”
This distinction between kineses and energeiai, Aristotle thinks, is not only crucial for thinking about the sorts of things that we do--such as opening mayonnaise jars or dancing--but for thinking about the sorts of things that we are. What are we? Well, at the most general level, we are what Aristotle calls “ousiai.” This term usually gets translated as “substances,” but that term doesn’t actually fit the Greek sense of “ousiai” very well, and I’ll take a bit of liberty here and translate here as “be-ers.” We are be-ers. Being is what do, and, in doing it, we are, and we are the sorts of things that we are.
How could being be something we do? It might seem that you don’t have to do very much in order to be. However, according to Aristotle, you kind of do. To see this, for our own case, we have to consider the particular kind of be-ers that we are. We’re living things. What a living thing does is live, and, in doing that, it is and is what it is: a living thing. Living is an activity, and it is through the doing of this activity that a living thing is. Now, the crucial point here is that living is not like opening a mayonnaise jar or climbing down the courthouse steps. It's not directed at any end external to itself. A living thing does do all sorts of things that are directed ends external to the doings of those things---for instance, or opening a mayonnaise jar or climbing down the courthouse steps---but living itself is not one of those things. The activity of living is an end in itself. That is to say, it is energeia, not a kinesis. For us be-ers, being (where, for us, this is living), is like dancing, not like opening a jar.
Mr. Meeseeks and Us
Let us now turn back to Mr. Meeseeks. Mr. Meeseeks is not like us. He is alive, but his living is not like our living. For Mr. Meeseeks, living is not an end in itself. Mr. Meeseeks lives in order to accomplish an end that is given to him from outside. So, when Mr. Meeseeks comes into being upon Rick’s pressing the button and is told to by Rick to open Jerry’s stupid mayonnaise jar, Mr. Meeseeks has the end of opening the jar, and his whole life is directed towards accomplishing this end. From the point that he is given that end, the Meeseeks lives to accomplish that end, an end which is not the living, but the getting done of something that is given to him. So, for a Meeseeks, living is not an energeia, but a kinesis; it is an activity that is directed towards an end other than itself. Indeed, Mr. Meeseeks it the very personification of a kinesis.
We said, a kinesis is an activity that aims, in being the very sort of activity that it is, to cease to be. Opening the jar is an activity that has, as its end, having opened the jar, and, once the jar has been opened, one is no longer opening it. So the activity of opening the jar aims to cease to be. Insofar as what it is for a Meeseeks to be is for it to live, where living is not an energeia but a kinesis, a Meeseeks aims, in being a Meeseeks, to die. This is, of course, just what they say:
"I can’t take it anymore! I just wanna die!"
“We all wanna die! We’re Meeseeks!”
Now, the standard view, I take it, is that the reason Mr. Meeseeks says this is that he is in pain and wants that pain to be no more. But I think that this is to make Mr. Meeseeks seem more like us than he really is. Mr. Meeseeks is a fundamentally different sort of being than us. That is to say, what being is for Mr. Meeseeks is distinct from what being is for us. For us, being is fundamentally an energeia, whereas, for Mr. Meeseeks, it is fundamentally a kinesis.
There is, I want to say, a philosophical moral here. Often, we go through life, acting as if it's really a kinesis rather than an energeia, like climing down the courthouse steps rather than dancing. But we need to remind ourselves that it's not. As Alan Watts puts it:
“We thought of life by analogy with a journey, a pilgrimage, which had a serious purpose at the end, and the thing was to get to that end, success or whatever it is, maybe heaven after you’re dead. But we missed the point the whole way along. It was a musical thing and you were supposed to sing or to dance while the music was being played.”
Taking a mini-seminar with Ayeh Kosman on Aristotle's Metaphysics several years ago is what got me thinking about the philosophical issues here, and, if you really want to dive into the Aristotelian way of thinking about things briefly discussed here, I cannot recommend highly enough his book, The Activity of Being.