epistemaulogies

AI: Absent Introspection

For the past few days it has felt “off-topic” to write or think about anything other than the attacks on free speech in the U.S. But I don't have a framework – yet – for writing about that, and I do have a close-reading framework for thinking about the ways AI encourages us to think about and use it.

When I say AI I'm shorthanding for “GenAI,” or generative AI – take your pick of chatbots like ChatGPT or image generation systems like DALL-E. I'll be using examples primarily from ChatGPT because it's still the most well-known brand-name AI, but most of the underlying principles are common to all AI – yes, even your favorite, yes, even ChatGPT4 or 5, yes, even the ones that promise to be significantly “better” than previous models. This is a way to understand what the project of AI is and does, even if you aren't an advanced computer scientist.

Let's begin by setting the stage. In close reading of literture, I might introduce the conventions of the genre, or discuss relevant biographical details. Here, let's get on the same page about what GenAI is. And since I like to let the text speak for itself, here's ChatGPT describing its own epistemology: How does it “know” things?

Generative AI is a predictive algorithm that uses extremely large (and no, not always “publicly available”) datasets as context to predict the next word in a sequence. By this laborious process – which is a clue as to why the technology is so incredibly resource-intensive – it constructs answers to queries that both trawl and curate available information and present it to the user in a facsimile of human speech. Rather than getting a list of relevant search results (and don't get me started on how “relevancy” in search results algorithms has been a slow and steady bait-and-switch) you are given a paragraph of “speech” as if a friend is explaining the concept to you.

This speech is prioritized and shaped by a training system. This part is very vague and heavily policed by the companies that control each AI, so it's tough to get specifics, but here is ChatGPT with a reasonable summary of how it works (I cut off the first part, which is essentially “crunching a big dataset on its own, by trying to predict a word and then checking against a source to adjust for loss”):

(I want to be clear, since it would be very silly of me, that I'm not taking ChatGPT at its word here. I've done my own research over the last few years on how GenAI works and this is consistent and sufficient for an investigation into the way its “knowledge” is both produced and professed to be “knowledge,” even if the picture is not technically precise. If you want to learn more about the computer science of it all, I recommend Emily Bender's work on computational linguistics.)

It's natural to pause here and consider: How did the companies, and the individual computer scientists, who developed GenAI decide on this approach? In what ways is it a natural outgrowth of their access to, and immense pressure to monetize, large datasets? How does it reflect a “computer science epistemology,” that is, a way of understanding the production of knowledge that is informed by the methods and the deemed-important questions of the field of computer science? And how did this all come to be labeled “intelligence”?

I encourage you to think more about these questions, and would love to hear your thoughts. In the meantime, let's keep focus on just the process by which AI produces “knowledge.” By predicting the next word in a sentence – and let's acknowledge that of course its analytical process is more contextual than that, so it's really predicting the next word in a sentence within a paragraph within a query that points it to a knowledge-base – it is necessarily a synthesis-machine. It will take what it can analyze and attempt to find a word – read, an answer – of best fit, defined as a combination of most-popular (the parameters set by its first round of training) and most-desired (the parameters of its second round of human-mediated training – who are the reinforcers and what is their epistemology, by the way?).

It is a predictor of the middle-way and then, worse, it parses that middle way into approximate, predictive text. And this is a best-case scenario, if it works perfectly – even though, as a necessary consequence of a predictive process, hallucinations are going to be a feature and not a bug in the system.

Does that process describe critical thinking? Does that process align with your own ideas of what it means to produce knowledge? Sometimes, when what you want is not knowledge but an approximation of general feeling, that might be a reasonable way to go about it – though, of course, you're allowing a dangerous distance between your own ideas of what “approximation” and “feeling” mean for veracity and what the AI presents to you. But in its own way, that's not necessarily harmful, if we do as we are told and “double-check important info.” However, in the next post I will discuss how the rest of the apparatus in these screenshots, not to mention the way AI is discussed and economically incentivized, encourages the user to use the tool in place of research and critical thinking – in place of the user's intelligence.

Thanks for listening. ~

Can technology be “read”?

I want to try out a concept that has been on my mind since college, when I had an idea for my first manuscript (back when it was obvious that I'd write and publish at least a handful, immediately and without angst): close-reading technology.*

Just as the form of literature encourages you to read and interpret it a certain way, just as the delivery and pretext of a speech provides auxiliary meaning, so too does the apparatus and context of technology encourage you to deploy it in particular manner.

So how does one “close read” technology? In literature and poetry, I might close read by selecting specific passages from the text and analyzing their form. For example, these lines from The Lake Isle of Innisfree by Yeats:

I will arise and go now, for always night and day

I hear lake water lapping with low sounds by the shore;

This line creates a rhythmic cadence using dactyls, a foot with a stressed syllable followed by two unstressed syllables. These dactyls, which are out of place in the mostly iambic poem, read more unevenly than the iambic feet. Placed within the only enjambment in the poem, the uneven dactyls contribute to giving the lengthy lines a disordered feel. The length of the line, as well as the way the dactyls seem to trip forward when read, makes the thought come off as a sort of ramble on behalf of the poet speaker. It is as if he is lost in the thoughts of the lake water, like they are haunting him.

Similarly, one might “close read” a hammer by examining elements of its form and design to determine how it argues it should be used. The long handle and weighted tip suggests – arguably demands – that it be used to swing, generating centrifugal force. The grip on the handle teaches you where to hold it for “best” performance. The blunt tip implies striking, not piercing. You see what I mean.

Why close reading, of all things? Close reading offers a schema for making the implicit explicit, particularly at the level of form – and here I’ll also point to my first post on Corey Robin’s approach to understanding conservative epistemologies.

Technology is typically thought of as a neutral tool, to be wielded for good or evil (but usually good), and also as a force that naturally evolves towards efficiency (for whom?) and benefit (to whom)? This has always been the contention – it is what the Luddites were revolting against, and consider how we view the Luddites. (Do you know anything about their movement besides that their name is a derogatory epithet for the technology-adverse? I didn't, until I specifically sought out a monograph on their history!)

Luckily, the GenAI movement has made this distinction extremely clear, and a significant number of us find ourselves more aligned with the Luddites than with Silicon Valley. Take, for example, these recent headlines about AI:

Fix The Risk, Don’t Ban The Tool: How To Secure GenAI At Work

5 myths holding back GenAI in the workplace

GenAI Complacency: The Costly Inaction in the Nordics

1 in 5 workers are misusing GenAI, according to a new survey

Notice the emphasis on moving forward (“costly complacency”, “holding back”, “secure”; there is a proper way to use GenAI and some people are “misusing” it, which could be you) and assumption that GenAI will become more widespread, whether you, the reader or professional, want it to or not.

On the other hand, there is a strong desire to cut through the hype and analyze what, exactly, GenAI can do, and what it means to use it:

GenAI in healthcare brings the need for risk policies

Do your students know the consequences of AI use on an internship programme?

GenAI — friend or foe?

I agree. I once wanted to close-read the internet (and I may get to that) but close-reading GenAI platforms is even more fruitful. Next post, I'll get into the details.

Thanks for listening. ~

*As an aside, I vividly remember wandering campus on a cold April night, feeling angsty the way 21-year-olds about to graduate do, wondering what I should do with my life. In a flash of insight I can only describe as devastating naiveté, I thought: I just need to decide what I would regret not doing, and then do that. Ipso facto, I will never have regrets; and once I accomplish that thing, I can live the rest of my life in tranquility. Were it so easy as being 21 makes it seem!

Breaking the Logic of the Gun

Disclaimer: This is a thought experiment, not a metaphor. It is not a suggestion. It is not a commentary. Any resemblance to real life is unintentional but not coincidental. It is a reflection of the explanatory power the thought experiment has for you.

There is a man holding a loaded gun to your head. He says you mean him harm and will never rest in your pursuit to ruin his life, so he has decided you have to die. He cannot be talked out of this by you, because he knows you lie, especially to him about your intentions for him. There is a party of neutral observers. They only know what he has told them. They are a society that does not generally tolerate murder, but they do think it can be justified. They are unarmed.

What can you do?

One: Disarm the man, but do not shoot. Let's say you are successful at disarming the man and now you have the gun pointed at HIS head, but you decide not to shoot. What is a safe next move for you? If he lives, he will pursue his agenda to kill you, because nothing has changed and now he knows you are even more powerful than he thought. The observers now either see you as a threat or as a person capable of defending themselves. Do you spend the rest of your life hunted? Do you think you can convince the observers that you are not a threat to this person? How?

Two: You disarm the man and shoot. Let's say you disarm the man. Now you have a gun to HIS head and, seeing that he will not stop in his pursuit of you, you decide to shoot and kill him. He is no longer a threat to your life, but you have become what he said you would become. The observers, even if some of them understand you might have acted in self-defense and that they would do the same thing, do not condone murder. Can you convince them you were forced into this? What about his followers, those who believed, and are now vindicated, that you are a threat? Are you irrevocably changed nonetheless?

Three: You die, protesting your innocence. You attempt, and fail, to talk down the man threatening you. You speak to the observers of your innocence. You are shot. Some of the observers believed you, and others just can't condone murder even if they aren't sure whether you are a threat, so they arrest the man. You are dead, so what does it matter to you?

Four: You die, accepting your fate. You do not protest your innocence. In a feat of extreme empathy, you tell the man you understand the position he's in and you don't begrudge what he has to do. It's very moving, but the man cannot be moved or persuaded. You are shot. The observers largely take your side because your surrender cannot be read as anything other than peace. They arrest the man and potentially learn a lesson. You are dead, so what does it matter to you?

Five: You run away. Let's be honest, you would be shot while running away. Again, perhaps this makes you a sympathetic victim, but you are dead. Best-case scenario, if you do get away, how do you spend the rest of your life as a hunted person?

Each time I think about this scenario, I'm struck by how futile the conflict between hunter and hunted becomes. You cannot escape the logic of the gun. You must either submit to it as a victim or as a victimizer, which marks you as a target for future violence. (I may write more on the cycle of violence.)

At this point you may be thinking, “Why are you skipping over the diplomatic option? Why is the man ontologically incapable of being persuaded?” First, I want to think through specifically the logic of violence. Second, consider: Whether the man is acting sincerely (in his belief that you are a danger to him) or cynically (because he stands to gain from your removal), what incentive does he have to listen to you? How difficult does his life become – especially after taking the step of publicly threatening you – if he backs down? How much would it threaten his own self-understanding?

What I keep coming back to in this thought experiment is the importance of the observers. Let's play out a few more scenarios:

Six: You appeal to the observers, and they step in. You are able to speak so eloquently that this neutral audience, who are not yet convinced of your badness, intervene and are able to disarm and safely contain the man. This is a good outcome, but consider: How are a population of unarmed observers disarming an armed man? How can he be contained in a way that ensures your safety and also breaks the cycle of violence (does not visit violence upon his person)? How can you stay safe from any of his followers who also believe you are a danger? How will they view this community intervention on your behalf?

Seven: The observers react strongly to the presence and threat of the gun, and intervene before you say anything.

This is currently my best formulation of this thought experiment. It breaks the logic of the gun by making the logic of the gun untenable and removing blame. It requires a uniform response. And don't forget that the gun here is a shorthand for all violence – how do we so successfully react and neutralize all violence? Still, there are degrees of relevance here – after all, this is essentially the current political response to nuclear threat, and that is an important peacekeeping logic that has certainly saved lives (perhaps all of our lives).

Can you think of any alternate scenarios? How does this experiment expand or contradict your thinking? Consider fantastical angles – what would it mean for the logic if you could become invincible to gunfire? How would that upend the man's thinking? Or that of the observers? What if you could die but come back to life?

Thanks for listening. ~

The Münchhausen Trilemma

I want to start by saying we are in a very gaslighting era. Agreed-upon “truths” have been replaced with Colbert-style “truthiness” driven, in large part, by what I described in my earlier post, this dogmatic epistemology that posits whatever the rich and powerful say is true because they are still rich and powerful. So I really don't mean the timing of this next idea to deepen that sense of “we can never really know anything.” Bear with me!

That said... how do you know what you know? How often do you sit down and think about how the precepts you trust made it into your own personal hall of knowledge? I think about it all the time, and the thing that made me tiresome is the Münchhausen Trilemma, also known as Agrippa's Principles (though those include two more “modes” which I don't really find useful).

The Münchhausen trilemma describes all the ways philosophers, specifically epistemologists, can think of to “prove” truths. When asked if something is true, generally someone will give evidence. But then it becomes reasonable to question THAT evidence, and the evidence for THAT evidence, etc. I think of this process like being a two-year-old: But why? But why?

The three ways to resolve the endless chain of questioning – parents, you're probably familiar with this – are circular reasoning (A is true because B, and B is true because A), ad infinitum (infinite) regression (A is true because B, B is true because C, etc. etc.), or a dogmatic assertion (“because!”).

None of those conclusions are typically how we think a “truth” should be established; in fact, generally each of those conclusions is considered a fault in reasoning. That's why the trilemma is named after German folklore of a man pulling himself out of a quagmire by his own bootstraps (itself a common fallacy nowadays). Of course, the secret fourth conclusion is simply that we don't know, and therefore the Münchhausen trilemma can mean one of two things:

1) We can't know anything; 2) The justifications we have for truths are logically unsatisfying.

Even though it seems like annoying philosophers prefer the first option, most epistemology is built off of sects who basically take the second statement as a starting point and then argue that the issues with their chosen solution – the infinite line, the loop, or the axiom – are trivial or can be solved or minimized by mental gymnastics.

For me, the trilemma has led me down three related paths of thought: Sophism (truth is agreed-upon?), dualism (our inability to resolve Truth stems from, or mirrors, our struggles to resolve the material world from an immaterial mind?), and deconstruction (all knowledge is metaphor?). Hence, many of my most annoying traits.

When I first learned of the trilemma I remember trying it with 2+2=4 (a classic) and the existence of atoms. I remember being struck not only that some of these foundational Truths were indeed infinitely reductive – or reductive until it reached an axiom – but also that I often couldn't regress more than a few principles back before I landed on, “I read this somewhere,” or “it is generally accepted.” Now I won't be studying mathematics to get to the axiom at the end of 2+2=4, but this process has made me attuned to the production of knowledge in each field of thought, which is honestly a fascinating and endlessly curious way to look at the world. (Someday I'll write about Kuhn's The Structure of Scientific Revolution as an excellent application of this curiosity.)

I neither take the trilemma seriously enough to be depressed nor lightly enough to be ignored. Instead, I think it's the most fruitful entry point to examining our own acquisition of knowledge, and from there the production of knowledge from without. By using the trilemma as a simple thought guide, you can start to easily see how much knowledge is persuasion, relativism, or simple dogma. And that's OK – it needn't lead to paralysis. Instead, I think it's actually made me more resilient to times like these where a comfortable common sense has given way to a struggle for ideological supremacy. And neither does it have to mean I don't “believe” in science, or trust physics, or question history – we'll get back to dualism (and I'll write about coherentism).

Try it yourself. Pick a truth, something that feels so fundamental and obvious that it can't be denied, and with applications to your lived experience. Then channel your inner toddler and ask: But why? Until you can't any longer. Where did you land? How did it make you feel? Does it change how you will think?

Thanks for listening.

The Battlefield and the Market

For the last year I, like many others, have become a little obsessed with the logic of the Trump voter. Family members become unreachable and dismissive; the profile of a voter voting against her interests, sometimes knowingly so; the newly political driven by a strange new animus. Pretty often this behavior is described as hypocritical, but I think, even as we say it, we know that doesn't find the mark. Hypocritical belies a kind of knowing subversion, a set of secret rules for thee and not for me. If true it invites shame or denial. But your average U.S. conservative has been placidly weathering these accusations for decades, to say nothing of the escalation of Trumpian rhetoric. “Yes,” they seem to say. “Rules for thee and not for me. What about it?”

In pursuit of this windmill I've started reading The Reactionary Mind by Corey Robin, the kind of political pop theory staid materialists don't usually read. His method for explaining the conservative logos is to close read a selection of reactionary writers and declaimers for clues into underlying forms of thought. Even if I don't think the methodology approaches an understanding of HOW reactionaries are motivated – after all, what describes the general pattern of reactionary movements? how do these ideas spread or renew and why in THOSE populations? – it's remarkably good at defining the problem.

Since the re-election I've been toying with Frank Wilhoit's definition of conservatism, which reflects some of the examples above: “Conservatism consists of exactly one proposition, to wit: There must be in-groups whom the law protects but does not bind, alongside out-groups whom the law binds but does not protect.”

You'll recognize in there some key characteristics of fascism, too, with the out- vs. in-groups. And I do think this definition is a good Occam's Razor for determining whether a policy or proposal is fundamentally conservative. But it still doesn't describe HOW reactionaries acquire and sedately defend this mindset. It's not in their “own words” and even if their logic follows this contour, it's not their logos.

In nearly a throwaway paragraph, which I think mainly functioned to conclude the first chapter on a grand theoretic and set up the next few chapters on Edmund Burke, Corey Robin provides a much more interesting paradigm. He writes:

“Unlike the feudal past, where power was presumed and privilege inherited, the conservative future envisions a world where power is demonstrated and privilege earned; not in the antiseptic and anodyne halls of the meritocracy, where admission is readily secured... but in the arduous struggle for supremacy. ... The battlefield is the natural proving ground of superiority; there, it is only the soldier, with his wits and weapon, who determines his standing in the world. With time, however, the conservative would find another proving ground in the marketplace.”

Then quoting William Graham Sumner to support his reading: “Liberty is a conquest.” And Burke: “At every step of my progress in life (for in every step was I traversed and opposed), and at every turnpike I met, I was obliged to shew my passport, and again and again to prove my sole title to the honour of being useful to my Country... Otherwise, no rank, no toleration even, for me.”

Now this – this is fruitful epistemology. I was immediately struck, taking notes, of the foreclosure:

If the battlefield/marketplace is the natural proving ground of superiority for the conservative it's a recognition of the tyranny of the victor in the construction of history – that's their epistemology. That which is true wins. That which wins is true. That which is false loses and therefore ceases to exist. It is anti-reflection. (Anti-intellectual.) It is purposefully action without thought.

The conceit is an anti-logos. Rather than reflective or persuasive argument, the conservative posits a worldview and waits for it to be challenged. In the very fact of still being in power, he is proven correct. Logical, ethical, or emotional challenges are walled outside of the arena, the effect like picketing gladiatorial combat then declaring “Roma victrix.”

And doesn't that just explain how comforting it can be! To not have anxiety over being right – you will either be right or you will be ousted. If you are not ousted – if you still have political and economic power – you are right. (Convenient, too.) Hence the anti-intellectualism, even though plenty of reactionaries are well-schooled, or persuasive, or interested in philosophy; that is simply not how the victor is decided, and therefore it is moot. Thus, too, why Trump and Musk are heralded thinkers despite objective measures; they are not (yet) defeated. They have power, and therefore truth.

The battlefield and the marketplace explains the allure of reactionary thought, it's imperviousness to accusations of hypocrisy (which are so devastating to the internal coherence of liberal thought), and its sneering imperiousness. It's remarkably animalistic, and in fact imbued with the same backwards reading of natural selection that many reactionaries employ to be racist: It persists, therefore it must be fit. It is an epistemology consistent with, and therefore able to solve the cognitive dissonance of, its inequal and unfree policy; and both are a reflection and a reading of the material circumstances that encourage this line of thinking.

At the very least, it gives me a schema for continuing to avoid Facebook arguments with Trumppilled relatives.

More, later, on the Münchhausen trilemma, how these views are propagated (why and how does the son of liberal parents become reactionary?), and maybe what can be done about it.

Thanks for listening.