This is a new format I'm trying out: short essays. The topics will be more diverse than my standard technical articles. They also allow me to write about topics that may not justify a full article, or in which I may not have the expertise to write a full article, but I still feel there's something interesting that needs to be said.
This comes from a Reddit comment I wrote recently. I thought the topic was interesting enough to have it in an article and expand upon it.
In this essay I'll try to explain why I struggle to see the value of a non-thesis CS master's in the US, even at a top school. I should point out that I haven't been through a master's program, but I have taken graduate courses, which are the same for master's students. Since courses is mostly what you do in a non-thesis master's, I think I have a somewhat good grasp of what the experience might be.
First, I assume that whoever is doing a non-thesis master's is not interested in research. This is because generally no thesis means no research, or at least no good research (e.g., you'll be at best a 5th author in some paper you didn't choose). At the same time, I don't think any research lab cares about a non-thesis master's, especially in the US where PhDs are long and advisors like undergrads to have time to train them. In short, you don't do good research or increase your prospects of doing that in the future. Thus, it seems obvious to me that a non-thesis master's is useless for anyone interested in research.
One other reason to do a master's is for the knowledge, and since you don't do any great research, this must primarily come from coursework. However, most grad courses (i.e., 500-level) do not consist of lectures by the professor. Instead, they mostly have student presentations. Unfortunately, most of these students don't give good presentations. This is either because they don't care or because they don't know how to, which is totally understandable because e.g., most are master's students straight from undergrad. So, you're just sitting there for one hour, listening to a not-so-amazing presentation (of an otherwise great paper), while your time would be much better spent: (1) reading the paper yourself, (2) watching the presentation from the authors.
Of course, there are exceptions. For example the Distributed Storage Systems course at UIUC is absolute fire (!) and I'm not even in systems. Almost all the sessions
were lectures by the advisor, and he went pretty deep on the topics covered. But such graduate courses are
rare! So, it just seems absurd to pay this insane amount of money to listen to
people giving their n-th presentation, n < 5, semester-in, semester-out.
So, if you get low-to-no-quality research experience and low-quality knowledge, what is left? Well, you could do it for the credentials. The problem here is that computer science is such that you can learn, experiment, and prove your value without needing access to e.g., specialized lab equipment. Also, you don't need any special qualifications to enter communities that allow you to do all this. More specifically, there is open-source software! First, I think most senior engineers would agree that open-source experience is one of the best things you can have on your CV. I can confirm that myself. But also, consider this: you're hiring for, say, an LLVM engineer. Why would you hire e.g., someone who took a grad course in compilers instead of someone who's been working on LLVM for the past few months?
Consider here that LLVM is an open-source and pretty welcoming community (something I've also experienced), so you can get this experience if you're willing to put in the effort. At the same time, working on LLVM requires basically just a laptop. And of course, the return-of-investment in terms of knowledge and experience of working on LLVM will be much higher than grad coursework. Simply consider that one is real useful engineering, while the other is, well... just coursework. For the same reason, it's also way more fun. As a bonus, you'll get to talk with some of the best engineers just by virtue of working on things together.
In short, open source is clearly the better option in terms of knowledge, experience, enjoyment, credentials, and connections. And it's free! So, I still don't see the point of doing a master's.
In my opinion, one should try industry directly. Ideally you'll land a job where you can do open-source as part of your job. But if not, do it on the side until you get a job you like.
Finally, I should note that I emphasized the non-thesis part because a thesis master's is a totally different story. If you care about research, a thesis master's can give you real research experience, papers, and research positions afterwards.
Descartes had a conception of infinity that seems much different than our contemporary understanding of the matter. This is kind of a big deal if you consider that infinity was central to Descartes philosophical arguments. It's also interesting because Descartes was not only a philosopher, but also a great mathematician.1 Yet, what we take today for granted was terribly elusive back then, even to the greatest minds. This “discrepancy”, among other things, shows how great the contribution of Cantor was; way beyond mathematics.
Let's delve into the matter. First, note that Descartes uses the idea of infinity to prove the existence of God (who, for Descartes, is an “infinite substance”). So, there's quite a bit at stake here. Let's try to understand what Descartes means by infinity. He never seems to address it head on, but in VII:47 he writes:
[...] even if my knowledge always increases more and more, I recognize that it will never be infinite, since it will never reach the point where it is not capable of further increase.
Descartes seems to confuse “infinite” with “inexpansible”. This directly contradicts how we understand infinity today. For example, take the set of natural numbers (ℕ). Clearly, ℕ is infinite. It is also clearly expansible. For example, we can expand it by adding to it elements from the set of irrational numbers (ℝ − ℚ).2 Not only that, but we can do that forever. In other words, it's not just expansible, but infinitely expansible. But still, it will always be smaller than the set of real numbers (ℝ).
What implications does this have for Descartes' argument? To be honest, I'm not really sure. But, the following seems somewhat clear to me: our conception of infinity would change, at the least, Descartes' Ontology. In short, an ontology is a categorization of concepts or entities that shows their properties and the relations between them. Without delving too deep, Descartes split everything that exists in the world into three types of substances: infinite substances, finite substances, and “modes” or qualities of substances. For example, we are a finite substance3, and an idea is a mode of a thinking substance (like Descartes).
Descartes grouped all things infinite into one category: infinite substances. But, imagine Descartes thinking, like us, about multiple types of infinity. For example, today we think that ℝ is a larger infinity than ℕ, even though both are infinite. Would that change Descartes' grouping? For example, would that create multiple levels of infinite substances? Maybe Descartes would argue that God is the largest possible infinity. But then here's the issue: Clearly we don't have our contemporary understanding of infinity because it exists somewhere inside us. Cantor had to come and do a lot of work to clarify this idea. It's definitely not one of the ideas that are “clear and distinct” for Descartes. So, it seems to me that would invalidate Descartes' argument that God is not a synthetic idea, created by us.
Tor Nørretranders writes in The User Illusion:4
The goal of science has always been to draw up the most concise description of the world possible. But there must be limits as to how concisely the world can be described.
Then, maybe formulating a unified theory is a type of compression. And maybe, then, there's a limit set by information theory. But, there's a caveat. It's not just about how concise a description is (i.e., how few bits it requires), but also how many phenomena it can describe/explain/represent. So, we have this problem where we don't care about e.g., a theory's text itself, but about how many things it represents. The genius in Maxwell's equations is not just in that they take up so little space, but mostly in that they describe so many things.
This indirection is what makes things hard and elusive. Because let's say we have the phrase: “the set of natural numbers”. Compression helps us answer the question: how can I represent this sequence of letters in the fewest bits possible? But, in the case of unified theories, we also care about the things a description (i.e., a formula, which would be the analogous of a phrase) represents; in this analogy, the phrase above represents the set of natural numbers. And this is why it becomes hard to formulate the problem in a way that we can attack it with compression tooling. It also showcases the magnificence of theories. Unified theories in Physics usually reduce the representation while increasing the concepts represented.
In this essay I will present an alternative interpretation of the recent rise of authoritarian regimes. I will first start by what I consider to be the popular interpretation, which as far as I can tell is based on Michael Sandel's work. This interpretation is founded on populism, according to which a small meritocratic elite has occupied most positions of power. The alternative interpretation is that of Theofanis Tasis, and it's a much less known5 but thought-provoking. He argues that new challenges posed by technology give authoritarians the opportunity to promise a return to a glorified past.
Disclaimer: I'm not interested in arguing here whether authoritarianism is on the rise; I just assume it. But if you don't think so, this essay may not be that interesting to you. I equally don't care about arguing whether authoritarianism is good or not. I simply don't find these political discussions that interesting.
The central body of work that analyzes the first interpretation is the book “Tyranny of Merit” by Michael Sandel. In this book, Sandel's narrative of what led to the rise of authoritarianism is founded on an instance of the populist framework. Because populism is one of these buzzwords that are thrown around carelessly, without necessarily most of us having any clear definition of it, let's start by defining populism. The definition I like the most is that of Cas Mudde:6
We define populism as a thin-centered ideology that considers society to be ultimately separated into two homogeneous and antagonistic camps, “the pure people” versus “the corrupt elite”, and which argues that politics should be an expression of the volonté générale (general will) of the people.
Let's briefly discuss what the terms "ideology" and "thin-centered" mean. Mudde has more comprehensive explanations of both; here, I will simplify and/or abridge his explanations. An "ideology" is simply a conception of the nature of people and the organization of society. In other words, “it is a view of how the world is and should be”. A “thin-centered” ideology is simply one that doesn't stand on its own—it's not full—but needs to attach itself to some other ideology (e.g., fascism, socialism, etc.). “Volonté générale” is the general will of the people. I will not go deeper into “volonté générale” because it has such a long history which I could never do justice to in such a short article. However, if you're interested, you can start your exploration from the Declaration of Human Rights and Rousseau's work7.
The reason I called this a framework is because you can think of it as a recipe—an abstract model with abstract variables.8 From that, you can create concrete instances by plugging concrete “values” for the variables; in this case, the two camps:“the pure people” and “the corrupt elite”. For example, one instance is that of Marxism, in which “the pure people” form the proletariat and “the corrupt elite” the bourgeoisie.
Let's see Sandel's instance in detail. For Sandel, the distribution to camps is done according to merit, and according to Sandel, we've come to believe that the metric for merit is formal education. In other words, the greater your degree, the greater your merit. So, “the pure people” are those lacking extensive formal education, e.g., those with just a high-school diploma. “The corrupt elite” is the set of people with college degrees. Sandel argues that in America's meritocracy, the positions of power have been occupied by “the corrupt elite”. With ⅔ of the U.S. without a college degree, most common people are left out of these positions, especially when we consider that the number of people in e.g., Congress, that don't have a college degree can be counted using one hand.
The problem is exacerbated if we consider that it's not just about who takes action. To understand that better, let's consider the case of a lawyer and her client. The lawyer is the technocrat who understands the legalese, the jargon that is needed to participate in legal processes, and the one who ultimately takes action. However, the lawyer is only a representative of the client in the true sense of the word in that she ultimately works for the benefit of her client. In doing that, and because the client should participate in going after their benefit, the lawyer tries to simplify, clarify, explain and dejargonize. But “the corrupt elite”—the technocrats—now implicitly (or even explicitly) tell the common folk that “you cannot even understand all that is involved; leave it to us”. They “represent” the public, but without the public's awareness of what's going on.
Thus, we arrive at Sandel's interpretation. Sandel locates the rise of authoritarianism in the promise that it will give back the power to common people. In other words, authoritarians have been emphasizing that if they come to power, they will give “the pure people” their voice back!9
Let me first do a small introduction to Theofanis Tasis' work. Tasis is a philosophy professor and most of his (recent) work concerns our conception of life10 in a society that increasingly adopts more technological advancements, and more specifically artificial intelligence. Unfortunately, most of his work is in Greek so I don't have anything to cite in English. But, I will do my best to transfer his work into a digestible explanation.
Tasis argues that the current society is more and more “iconistic”, meaning that citizens live partly in the physical world, and partly in the digital world of social media. Because of the ease of global access to social media profiles, this leads to a fusion of the private and public spheres. This means that we have an image of our self that is partly physical and private—for example, when we gather up at a friend's house—and partly virtual11 and public—for example when we post something on TikTok.
This iconistic society is challenging because it requires new skills from its citizens. At minimum, citizens need to acquire the skills required to operate the necessary devices, and to understand the processes, conventions, and means provided by the virtual counterpart of our society and its platforms (e.g., Twitter, TikTok). Most importantly, however, one needs to learn how to operate effectively and efficiently in these virtual societies if they are to succeed. Success here means the usual: political power, money, influence, safety, etc.
But these societies and their processes are elusive. For example, we live in an attention economy and so to get ahead, one probably should grasp what can make them popular. The game is not simple and it's mediated. It's mediated by obscure algorithms that heavily impact who comes first and which orchestrate the attention economy.
Tasis argues that only a small elite possesses the knowledge that is necessary to get ahead; these would make up “the corrupt elite” in the populist framework. This leaves behind a large portion of the common people who don't possess these skills; those would make up “the pure people”. According to Tasis, then, authoritarian regimes arose through a promise of a return to a glorified, and mostly physical, past; a more traditional view of what it means to be a citizen.
For me, the fascinating question here is: to what degree are these two interpretations two faces of the same coin? In other words, how much does Sandel's corrupt elite overlap with those who can operate effectively in an iconistic society? On the face of it, they seem to be in agreement: those who can't operate technology effectively are mostly older generations, which also have fewer degrees.
At the same time, I think the two theories are in disagreement with regards to the recent political events in the U.S. In particular, (assuming that Trump is authoritarian12), many Trump voters are young; Gen-Z in fact. And these people clearly are more in tune with the iconistic society than e.g., old Democrat professors of Sociology. So, it seems that Tasis' theory doesn't explain these events convincingly. At the same time, Tasis' interpretation seems way more "up to date". Sandel's explanation seems to ignore many central dimensions that our lives have today, especially with the intrusive role of technology in general and artificial intelligence in particular.