I've seen a student using ChatGPT for almost everything in his PhD (in engineering). You have some data but you don't know what kind of statistical analysis tool to use? Ask ChatGPT. Code for the analysis? ChatGPT. How do you interpret the results? ChatGPT.
And so on. I could bet that some of his scientific questions where generated, and that's no surprise to me, it's just SO easy this way and if the PhD advisor just says "ok that's good" and no one ever complain during the PhD defense then for sure this will keep going.
But regarding OP's link, when Lemire says I kept my mouth shut. I am never rude on purpose. You can't say that and complain that academia "is blown to bits". It's your responsibility as a scientist to step up and say that some research is garbage when you see it.
> It's your responsibility as a scientist to step up and say that some research is garbage when you see it.
The only case I personally know of someone doing that during their PhD didn't end well.
My friend couldn't replicate the results from a known professor in the field, asked for the data + model to re-run because he assumed his own work was wrong and wanted to benchmark against the known study. Got stonewalled for more than a year, brought it up with supervisors because he started getting the feeling the results were tampered and the professor didn't want to be found out. He pushed it but got ridiculed by the professor's university ethics committee.
After a couple of years he could show that the research was at least sketchy and he depended on that model/results for his own work, he lost 2 years of research and completely left academia after finishing the PhD (delayed by almost 2 years).
Maybe it wasn’t garbage, maybe it’s just gotten significantly easier because the hard parts have gotten less hard. Search is solved, prose is solved, reasoning is… assisted at least. What he wasn’t being rude about was bursting his colleague’s bubble about AI generated work in general.
We’re at the point of learning that some things that used to matter no longer do, and some things we used to think mattered never did. It’s a going to be a shock to everyone, and this same phenomenon is happening everywhere, not just academia or software engineering.
My bet: a lot of the things we used to do were habit, ceremony and gatekeeping much more than being necessities.
You’re probably right, but I’d also bet some things that we’re about to sweep aside will turn out to have been necessities and we’ll find it very difficult to bring them back.
To be pedantic about it, if something is a necessity it can’t be done without by definition. So if you’re in that position whatever you’re talking about wasn’t a necessity. Is this like Socrates thinking that reading will make people lazy and forgetful?
More like how using hand-held computers programmed for dopamine hits could further entrench a sedentary lifestyle, reduce social cohesion, and lead to individual brain rot
> If something is a necessity it can’t be done without by definition
You fail at being pedantic. If you wanted to be pedantic about this, it would first be necessary to define what it is a necessity for. Are the things being swept aside a necessity for the universe to continue existing? No, definitely not. But perhaps they could be a necessity for the (academic) culture to prosper. Which do you think is more likely that the person you responded to meant, that they were talking about things being necessary for the universe to keep existing or things being necessary for our culture to flourish?
There is nothing worse than a half-assed pedant who can't even be sufficiently pedantic to make a point correctly. Really, that's closer to autism, not pedantry. There are inferred clauses when people speak, but you (choose to?) intentionally disregard them and interpret them in the wrong way, thereby arguing against things that weren't said. Even programs can infer unstated clauses by context (eg. can successfully infer from `var x = 2` that x is an int despite it not being said, without explicitly declaring `int x = 2`).
Wow, that’s needlessly aggressive, but you do you.
I wonder what it is about this subject that makes people so angry? I think I’ll settle on status. People are afraid of having their social status taken away by AI and become irate at the thought of it.
My response to you has no bearing whatsoever on the subject of this thread. I work at a successful AI startup and am set to retire for life comfortably no matter what happens. It's your specific class of action that irritates me. These orange websites are plagued by people intentionally misrepresenting other people's words and telling them they're wrong based on things they didn't actually say. It's the most annoying, detrimental-to-conversation kind of comment you can possibly make. And it does pedantry a bad name! There's nothing wrong with being pedantic, and drilling into the details to get everything right. But drilling into the details and getting them wrong is offensive and the only purpose it serves is to annoy everyone around you.
Going back to my programming example, suppose your compiler decided to interpret 'var x = 2' as a string and then told you you were wrong, throwing an error instead of compiling your program, even though it's the compiler being defective and not adhering to the established language rules for type inference. That's you, in this conversation. That does not add to the conversation. The compiler needs to be fixed so it stops doing that.
So the value is in those who where actually kept outside by the gatekeepers. Those that can use the new tools and are no longer kept at bay by those that need artificial moats. The very same innovation that creeps to the front of society in a war, is going to be liberated. I toast to that.
What is possible as ground-truth, if the "tarbabies" of society get sidelined can be seen in Ukraine.
I find that I'm much more tolerant of bad spelling and terrible punctuation because I know that a human being wrote it in the end. It's kind of hilarious that a sign of poor quality is now a sign of passion if not quality now.
Really depends on how the resource wars pan out. If inference continues to get monopolized and the monopolists do a rug pull in terms of availability of AI tools, the old ways will still matter.
When these things no longer matter, who will that benefit and who will be disadvantaged? What will the costs or net benefits to society be? Disruption can be good or bad, or often a mix. It's hard to say now whether social media was a net benefit, but it has definitely been disruptive.
The author is a mathematician and I think to a certain extend the tweet reflects the current panic among (some) mathematicians. So ~~second sentence~~ third paragraph the claim that somehow ai could right now write a phd in physics or sociology is something we don't observe (at the moment). What we observe is, that ai can find counter examples to well established conjectures in mathematics quite well, but the thing is the other fields don't have the kind of well established riddles that currently produce the flashy results in mathematics.
To show my ignorance in mathematics a bit: do you feel that having such neatly defined riddles gives the AI an advantage in solving them?
A lot of Innovations or insights are obvious in hindsight, but no one thought to consider the problem, and put the pieces of the solution together. In this sense a well defined problem is a large portion of the solution as well.
I mention this because I feel AI software agents have a huge advantage due the body of prior work available to them and how provable solutions can be. This I feel gives the impression that the agents are more generally intelligent than they actually are.
That's just lack of observation. Which PhD candidate is going to say "Chat wrote it for me?" We know a lot of academic articles are AI written. And a very, very large part of the student essays. Unless intercepted, they'll end up in the thesis. And in sociology, the texts are so vague, that it becomes even harder to pick out slop.
There are good reasons to assume PhD students see an advantage to using AI, so they will.
I am not sure what point the author is even trying to make here. On the one hand, he seems to complain about how AI has virtually made the traditionally PhD thesis obsolete, but on the other hand, he also states:
> Consider that most PhD theses were never good. How often do you rush to read a PhD thesis? The vast majority of them are painful to read. You learn little if anything.
So, it sounds like nothing of much value has been lost.
I think what he really complains about is that AI is starting to show that the emperor called academia has no clothes. So much of working your way through that system has always been about being able to master largely pointless rituals.
Yet, the people on the inside have no interest in making any improvements, because academia has always been institutionally conservative. But now AI is starting to put pressure on them to rethink their way of doing things, and they really don't like it.
Most thesis are never read because anything worth sharing with the wider world ( and some that isn't ) is highly likely to have been published as a paper - not because the work in thesis has no value.
The whole point of a PhD is not to create a thesis - that's just a mechanism to measure - it's to be trained as a scientist or researcher.
Doing a degree in chemistry for example, is largely a knowledge building phase - and in my view it doesn't make you a scientist - being a scientist is about discovering new things about the world that nobody else has - ever - that's what you are learning how to do when doing a PhD.
I think the premise was, before you had to actually do some work to create the thesis. And there was always the concern that yours could be one which was read deeply, so that thesis work had to at least show that you did some work. That there was meat behind the paper.
But now, it could simply be all a couple of prompts to an LLM.
The bar is just lower for not doing the work, now.
> So, it sounds like nothing of much value has been lost.
Do you actually mean this? To me this sounds like someone said "you never learn anything by reading a high school essay", and you reply "so stop writing them". The point is not the product, you obviously train people by making the product.
> AI is starting to show that the emperor called academia has no clothes.
Seriously, what are you talking about. Academia in the last century has been the most successful engine of knowledge and technology in human history. Lots of papers are junk, like lots of businesses are junk, lots of books are junk. But I don't know how any serious person can say academia has no clothes.
PhDs only were adopted universally in 1917 with some resistance and apprehension.
The issue here is that academia forgot what it was about a long time ago and is now having to face the consequences for a hundred years of bad decisions.
The largest output of a PhD has always been the training to the student, not the thesis itself, hence why we're called 'students'. Anyone claiming that a thesis can be generated by AI is missing the point. AI can also do everything an undergrad can do, we don't claim that undergrad education has been blown to bits. At best, we say we need better modes of evaluation, and perhaps that's true for PhDs as well.
I submitted my thesis this month at a QS top 10 uni after nearly 4 years of work. LLMs were available for most of that time. I don't really feel that it has diminished the value of my thesis by much really.
>> we don't claim that undergrad education has been blown to bits
> But it has been.
Let me fix it: pen and paper. There you go, bada-bing bada-bum. In our uni the exam for Algorithms and Data Structures (one that most people struggle with) is done on paper in an exam room. You better know your binary search.
People, especially wealthy, have been coasting through education paying someone to write their papers since Great Pumpkin knows how long. Now you can do same thing cheaply with LLM’s. The solution is pretty simple, let them fail. Have them show what they can do live in front of examiners. Here’s a computer without network card, only Python and SQLite installed with the Python basic documentation, build me X and prove it works in Z, Y, and Q.
I am studying my second degree and although there’s million and one ways to cheat, I won’t, since A) I want to learn this shot B) I am not sure LLM’s will be available with these capacity and with theses prices in the future, at least I won’t count on it.
Live exam, perfectly fine. You can use an LLM as support while you study, no problem, as a glorified search engine. On steroids. From the future. Then you prove you understood it.
Writing an essay, or a thesis, with LLM assist, is difficult to avoid in this scenario, though. It would require your prof, TA or examination committee to read the thing fully and understand the topic deeply. For 101 courses, that's doable, but for a PhD, that's so much work, that it'll break the system immediately.
Montaigne originally used essay (from French essayer, to try, attempt) to test hos on knowledge and understanding, his own capability to create new knowledge. In plain English, you test yourself if you can use your brain. That’s why it is so powerful to write a blog post about a topic you have just learned. Obviously, if you just generate the thing with LLM, there’s no point.
Essay is a fairly recent invention and perhaps it was not very good for testing students un the first place, since it kinda was not meant for that originally? So maybe we should test knowledge and understanding in different ways? In STEM-subjects this is not very hard, since basically you either can do the thing or not (like my example of coding above), but I have no idea how one would do this in history or philosophy. Should we go back to having long walks in the garden of the academy, where the professor guide’s you towards knowledge like some 21st Century Socrates? That doesn’t seem doable at this level of university students.
And how would you test PhD -level knowledge? You have to build a nuclear reactor from scratch in front of live audience? Or perhaps the oral examination have to become a real exam, not just a pointless ritual? Anyway, I think my point on using essay as a testing structure still stands. Maybe the examination should be like BJJ belt exam, where you have to show what you actually can do in front of peers and your teachers. A lot more people are likely to fail, but so be it. Maybe having a PhD stop being a joke then.
You’re sort of saying that’s because your thesis had no value to begin with, because it’s mainly a teaching exercise. I’m not saying that it doesn’t have value by the way, I’m sure it was good, that’s just my read of your post.
It’s an interesting subject. Makes me want to vibe code a PhD generator just like in the tweet. Maybe I will.
They are saying that the value of most PhDs was in the learning the student got in the process, not in the PhD itself. Groundbreaking PhDs exist, but they are far, far, from the norm or expectation. Exceptional people doing exceptional work will always exist.
> your thesis had no value to begin with, because it’s mainly a teaching exercise
And that’s the problem in your understanding. Thinking that a teaching exercise has no value while OP was saying that IS the value. It’s missing the forest for the trees. The point of homework isn’t to solve the problems. I bet you the teacher assigning the homework already knows the answers. Just like the point of a marathon isn’t to travel 26 miles because you could just take a bus.
Well no, it’s a problem in your understanding of what I’ve written. The thesis and what’s learned along the way are separate objects with separate value, and you’ve decided to misread me in order to have something to be indignant about. However, to expand on a point I didn’t originally raise: learning itself is likely also reduced with AI assistance. That’s a fairly natural consequence of having to do less work yourself, we don’t retain information we don’t need to retain.
The value of the exercises you do in school (from sums to PhDs) is the learning you get from them. Good, because you need people to learn stuff, because the people that know stuff will be dead in a few years and you need to replace them.
Except, if machines can do the exercises up to the PhDs, you probably don't need those brains in the future, at least not as many as before. You might need only the brightest ones, and who knows for how long? The next ten years?
Yes, that at least is a valid argument. I do’t really agree with it though. Humans need meaning and purpose in life. Intellectual pursuit is that meaning or purpose for a lot of people. The Educational institutes might change drastically from how they have been in the last few centuries, but people will never seize learning. The act of learning is in itself incredible rewardig to many people.
> at least not as many as before. You might need only the brightest ones, and who knows for how long? The next ten years?
Then those with power will keep it and pass it on to their heirs, and some without will try to finesse it but the masses will be told “we don’t need you”. I think in a lot of ways, this is similar to how feudal societies were. Maybe humanity will pass through another phase of that. Maybe we already are. But I think it won’t last either.
I think they were saying the "largest" output was the training, which I agree with. But it's also the case that they nudge forwards thinking in various areas through the outputs of the research process, whether that's papers, talks or the thesis itself.
> In the past, the path to a math Ph.D. cultivated resilience, resourcefulness, critical thinking, and a healthy skepticism (…) but suddenly you can produce a passable Ph.D. thesis with the push of a button
But actually nothing changed. Or maybe a path to "resilience, resourcefulness and critical thinking". Because human brains still needs to be shaped by years of training on some quality "literature", of some form.
We still must/want to human-[re]check important results, right ? And that require years of students time dedicated to memoizing facts and doing exercises in discovering already discovered results - learning and weights tuning, in the brains.
Yes, demotivator factor is very high or maybe just more visible then usual. Especially for brain paths forming - an that process is not quite stated in university and other education...
I think that when LLMs finish words and sentences shuffling and finds most of low hanging fruits in cutting edge of research ;) then only humans can move things forward, via abstractions, syntesis or old good paradigm abandoning. Hard to imagine LLM on their own "discover" something and then drops all that "literature" it was trained on as obsolote :) In next prompt it will happily return you old texts without any influence of just discovered paradigm shift.
In XIX century we got quite stagnation in science - it was belived that everything was already discovered, explained, just some few experiments are needed because some numbers do not adds up... And that proliferated to philophy and culture via some "proofs" for atheists. But in 1905 a paper was published... Too bad politicans do not get implications of that and still was pushing communism decades later...
So we realy want humans with brain pathways shaped mostly "old way" - the only one way available for human beings - by that training called "education". As always there is resistance and pain and attempts to find a shortcuts by cheating. Maybe this is time to clearly state that brain workings training is big part of education ? Just like in gym you are repeating to trying to lift weights up to your limit and even little above, with supervisor oversight.
Einstein - definitive end of XIX ilusion and way of thinking that "We know everything. We are the lords of things, we own all of it! And other peoples too! We can engineer societes as we like it!". With strict oposition to old and humbler "God" or natural ways. With aggresive "we" that do not pay attention to rights of other we's.
And, IMO, Einstein, Godel and more discoveries should be reason to not continue communism stupidity. Especially to not continue - after year 1905 - social engineering plans of "world-wide GENOCIDE until everybody is good citizen".
Of course I'm talking about best-wanting-for-humans but still naive communism and not about a way to destabilize other countries by any means necessary. Such move did almost insta-karma effect on Germany: sponsoring commies in Russia resulted in nazi being sponsored in Germany...
Don't think someone want to deny societes, philosophers and politicians are drived by overall science view on reality. And what else could drive atheistic minds ? Not judging here, logic is everyones duty.
Also my "too bad politicians..." was obviously expression of a wish. Not only science influents politicians. But discovered facts (counterarguments in this case) about reality could be taken for consideration, right ?
In my opinion, we will continue to need fully educated people who read books and peer-reviewed articles. We need literature, not "'literature', of some form."
“If algorithmic targeting disrupts your society, the culprit is most likely not algorithmic targeting”
Or replace AI/algorithmic targeting with tech in general and the disruption target with whatever it targets and you’d realize the problem with the sentence.
Technological advances have disrupted plenty of fields. That doesn’t mean those fields were fundamentally flawed. Every arena has a certain degree of dysfunction. AI has its own massive share of issues already. But that doesn’t negate the whole field.
Take the classic example of the Travel Agent. They are all but extinct because of technology. Yet they did serve a legitimate purpose before. Yes, plenty of them were middlemen who didn’t care, but also plenty were passionate about organizing travel plans and helping people arrange their travels and vacations. Plenty of people using AI today are also middlemen between you and Claude who also don’t care
If tech is disrupting X, that’s a problem with X” tends to be said by people who have X comfortably sorted themselves (e.g., already have a PhD that counts as it’s from the pre-LLM days).
In the end it depends on whether one considers technological progress or human flourishing to be the end goal. Surprise—they are not always aligned, and sometimes are at odds with each other.
Agreed. It can also be people who have problems with X for other reasons and like to see X “disrupted” (e.g. a disgruntled grad student who got screwed over by a bad advisor or a hard working and capable office employee who got passed over for a less capable, less hardworking PhD holder).
Like I had many bad experiences in cabs in my younger years. It was just bad luck really, but I hated cabs for the longest time because of that and completely dreaded needing them. When Uber was first coming on the scene and cab drivers were protesting against it, I did think to my self something along the lines of “if an app can destroy your business, the problem was in your business” but I really just didn’t like cabs for other reasons. And the irony is now Uber is the same. When I land in my airport, I see crazy lines for rideshare with people waiting on cars to come pick them up, while there is a cab on the other side that is ~$15 cheaper somehow and ready to go with no lines. And I get to tip the driver those saved 15 bucks
Travel agents are not extinct AT ALL. They pivoted from selling margin-less flights to design more curated travel packages. Different and nichier market for sure, but not extinct.
Fair point. They also still exist for corporate and business travels. But it’s fair to say the field did get massively disrupted (god I hate that word) because of technology.
In no way getting a PhD or doing published research is an act of science or human progress, and it's been like that for probably 20 years.
It's an act of coordinated methodology in showing respect to previous peers by acknowledging that you read what they wrote when they were acknowledging even more previous peers.
Every attempt to do something new is rejected unless you take 99% of something that has been done and try to add your 1% to it.
But the thing is that you don't even want to do that, you need to either have a publishable/defensible thesis, or if you already have a PhD, pursue a tenurable track and grants which never collides to productive human progress.
So yes, AI could very well run tenure track career more efficiently and with better results. It's not AI the problem: it is the checks and performance indicators that are in place that make it very easy for AI to dominate and very exhausting for a human to follow.
A good researcher with good AI knowledge would (and should) dominate their field.
* Medicine is luckily saved from this, with some exceptions.
Maybe old methodologies will be discarded and new ones will emerge.
Thinking back, when I first started teaching myself programming, I didn't know what to learn, so I explored the history of programming and organized it as I went.
One of the most striking things I remember is that when Stack Overflow first launched, quite a few people opposed it.
Also, I recall that in ancient Greece, Socrates criticized writing, saying it would weaken human memory.
When SO first appeared, there were many who insisted that the only proper programmer's way was to RTFM, deeply understand the system's fundamentals, and then write code. I wasn't from that generation—I belonged to the copy-paste-from-SO generation—so I can't say for sure, but I found it quite fascinating.
The cost of that friction could only be borne by a very small minority, and that minority could guarantee quality. That's why scholarship was something only the elite could pursue—and to some extent, it still is.
In the past, tasks were painful and high-friction. The results were filtered through that process, accessible only to the few who could endure it. But we tend to mistake those inefficient drops of sweat for quality. In reality, just as writing didn't diminish philosophy but rather created systems like law and philosophy, this might just be another turning point
I think Mr. Lemire's post is similar in spirit. In other words, when friction increases, the cost of production for producers also rises. So back then, everything produced through that high-friction process was easier to quality-control. But that's no longer the case.
And actually, universities were originally about 'holistic education,' but these days, they've become more about training talent for industry and managing human resources for the job market. That shift has caused problems.
In that sense, it's only natural that these problems arise at the intersection of academia and industry.
Industry usually demands 'people and technologies that can boost productivity right now.' Meanwhile, academia should ideally pursue problems worth exploring over the long term, even if they have no immediate utility. But the current state is a product of compromise.
Once university evaluations, student recruitment, research funding, and employment rates become tightly linked to industry demand, the latter starts to pressure the former. And under those conditions, the current outcome is almost inevitable—because industry increasingly wants to churn out degree stickers at lower and lower costs.
In the end, a different methodology will be needed, and whoever proposes it will become the game changer. Then new schools of thought and methodologies will emerge based on that person, and they'll gain enormous fame. I'm curious who that will be.
New things are always born by laying the past to rest. I'm always waiting for that new methodology.
>Industry usually demands 'people and technologies that can boost productivity right now.'
Industry demands a giant sorting machine and that is what they got. If we acknowledged and accepted this we could come up with a much better solution than what we have.
I've seen a student using ChatGPT for almost everything in his PhD (in engineering). You have some data but you don't know what kind of statistical analysis tool to use? Ask ChatGPT. Code for the analysis? ChatGPT. How do you interpret the results? ChatGPT.
And so on. I could bet that some of his scientific questions where generated, and that's no surprise to me, it's just SO easy this way and if the PhD advisor just says "ok that's good" and no one ever complain during the PhD defense then for sure this will keep going.
But regarding OP's link, when Lemire says I kept my mouth shut. I am never rude on purpose. You can't say that and complain that academia "is blown to bits". It's your responsibility as a scientist to step up and say that some research is garbage when you see it.
> It's your responsibility as a scientist to step up and say that some research is garbage when you see it.
The only case I personally know of someone doing that during their PhD didn't end well.
My friend couldn't replicate the results from a known professor in the field, asked for the data + model to re-run because he assumed his own work was wrong and wanted to benchmark against the known study. Got stonewalled for more than a year, brought it up with supervisors because he started getting the feeling the results were tampered and the professor didn't want to be found out. He pushed it but got ridiculed by the professor's university ethics committee.
After a couple of years he could show that the research was at least sketchy and he depended on that model/results for his own work, he lost 2 years of research and completely left academia after finishing the PhD (delayed by almost 2 years).
Maybe it wasn’t garbage, maybe it’s just gotten significantly easier because the hard parts have gotten less hard. Search is solved, prose is solved, reasoning is… assisted at least. What he wasn’t being rude about was bursting his colleague’s bubble about AI generated work in general.
We’re at the point of learning that some things that used to matter no longer do, and some things we used to think mattered never did. It’s a going to be a shock to everyone, and this same phenomenon is happening everywhere, not just academia or software engineering.
My bet: a lot of the things we used to do were habit, ceremony and gatekeeping much more than being necessities.
You’re probably right, but I’d also bet some things that we’re about to sweep aside will turn out to have been necessities and we’ll find it very difficult to bring them back.
To be pedantic about it, if something is a necessity it can’t be done without by definition. So if you’re in that position whatever you’re talking about wasn’t a necessity. Is this like Socrates thinking that reading will make people lazy and forgetful?
More like how using hand-held computers programmed for dopamine hits could further entrench a sedentary lifestyle, reduce social cohesion, and lead to individual brain rot
> If something is a necessity it can’t be done without by definition
You fail at being pedantic. If you wanted to be pedantic about this, it would first be necessary to define what it is a necessity for. Are the things being swept aside a necessity for the universe to continue existing? No, definitely not. But perhaps they could be a necessity for the (academic) culture to prosper. Which do you think is more likely that the person you responded to meant, that they were talking about things being necessary for the universe to keep existing or things being necessary for our culture to flourish?
There is nothing worse than a half-assed pedant who can't even be sufficiently pedantic to make a point correctly. Really, that's closer to autism, not pedantry. There are inferred clauses when people speak, but you (choose to?) intentionally disregard them and interpret them in the wrong way, thereby arguing against things that weren't said. Even programs can infer unstated clauses by context (eg. can successfully infer from `var x = 2` that x is an int despite it not being said, without explicitly declaring `int x = 2`).
Wow, that’s needlessly aggressive, but you do you.
I wonder what it is about this subject that makes people so angry? I think I’ll settle on status. People are afraid of having their social status taken away by AI and become irate at the thought of it.
My response to you has no bearing whatsoever on the subject of this thread. I work at a successful AI startup and am set to retire for life comfortably no matter what happens. It's your specific class of action that irritates me. These orange websites are plagued by people intentionally misrepresenting other people's words and telling them they're wrong based on things they didn't actually say. It's the most annoying, detrimental-to-conversation kind of comment you can possibly make. And it does pedantry a bad name! There's nothing wrong with being pedantic, and drilling into the details to get everything right. But drilling into the details and getting them wrong is offensive and the only purpose it serves is to annoy everyone around you.
Going back to my programming example, suppose your compiler decided to interpret 'var x = 2' as a string and then told you you were wrong, throwing an error instead of compiling your program, even though it's the compiler being defective and not adhering to the established language rules for type inference. That's you, in this conversation. That does not add to the conversation. The compiler needs to be fixed so it stops doing that.
Is that your way of saying: I don't care? Because it sounds are if you're relativizing the problems to the level where you don't have to.
You don't specify what you think "no longer matters." And so there's no way to judge what you think is the impact.
> My bet: a lot of the things we used to do were habit, ceremony and gatekeeping much more than being necessities.
Yeah, right. "I've got a calculator, so we no longer need to learn arithmetic." It's the way of the ostrich.
It’s a general sentiment about the way I think things are heading. You seem angry about it, that’s for you to figure out for yourself.
Nero probably even enjoyed seeing Rome burn. I think other people might have been angry. Don't be Nero.
So the value is in those who where actually kept outside by the gatekeepers. Those that can use the new tools and are no longer kept at bay by those that need artificial moats. The very same innovation that creeps to the front of society in a war, is going to be liberated. I toast to that.
What is possible as ground-truth, if the "tarbabies" of society get sidelined can be seen in Ukraine.
I find that I'm much more tolerant of bad spelling and terrible punctuation because I know that a human being wrote it in the end. It's kind of hilarious that a sign of poor quality is now a sign of passion if not quality now.
There's an LLM prompt that will trivially do that.
Really depends on how the resource wars pan out. If inference continues to get monopolized and the monopolists do a rug pull in terms of availability of AI tools, the old ways will still matter.
>It’s a going to be a shock to everyone..
Not to everyone though...
When these things no longer matter, who will that benefit and who will be disadvantaged? What will the costs or net benefits to society be? Disruption can be good or bad, or often a mix. It's hard to say now whether social media was a net benefit, but it has definitely been disruptive.
The author is a mathematician and I think to a certain extend the tweet reflects the current panic among (some) mathematicians. So ~~second sentence~~ third paragraph the claim that somehow ai could right now write a phd in physics or sociology is something we don't observe (at the moment). What we observe is, that ai can find counter examples to well established conjectures in mathematics quite well, but the thing is the other fields don't have the kind of well established riddles that currently produce the flashy results in mathematics.
To show my ignorance in mathematics a bit: do you feel that having such neatly defined riddles gives the AI an advantage in solving them?
A lot of Innovations or insights are obvious in hindsight, but no one thought to consider the problem, and put the pieces of the solution together. In this sense a well defined problem is a large portion of the solution as well.
I mention this because I feel AI software agents have a huge advantage due the body of prior work available to them and how provable solutions can be. This I feel gives the impression that the agents are more generally intelligent than they actually are.
Is this another example of that perhaps?
> something we don't observe
That's just lack of observation. Which PhD candidate is going to say "Chat wrote it for me?" We know a lot of academic articles are AI written. And a very, very large part of the student essays. Unless intercepted, they'll end up in the thesis. And in sociology, the texts are so vague, that it becomes even harder to pick out slop.
There are good reasons to assume PhD students see an advantage to using AI, so they will.
https://xcancel.com/lemire/status/2082851447499088173#m
I am not sure what point the author is even trying to make here. On the one hand, he seems to complain about how AI has virtually made the traditionally PhD thesis obsolete, but on the other hand, he also states:
> Consider that most PhD theses were never good. How often do you rush to read a PhD thesis? The vast majority of them are painful to read. You learn little if anything.
So, it sounds like nothing of much value has been lost.
I think what he really complains about is that AI is starting to show that the emperor called academia has no clothes. So much of working your way through that system has always been about being able to master largely pointless rituals.
Yet, the people on the inside have no interest in making any improvements, because academia has always been institutionally conservative. But now AI is starting to put pressure on them to rethink their way of doing things, and they really don't like it.
Most thesis are never read because anything worth sharing with the wider world ( and some that isn't ) is highly likely to have been published as a paper - not because the work in thesis has no value.
The whole point of a PhD is not to create a thesis - that's just a mechanism to measure - it's to be trained as a scientist or researcher.
Doing a degree in chemistry for example, is largely a knowledge building phase - and in my view it doesn't make you a scientist - being a scientist is about discovering new things about the world that nobody else has - ever - that's what you are learning how to do when doing a PhD.
A good place to plug my favorite thesis to read: Okasaki’s “purely functional data structures” https://www.cs.cmu.edu/~rwh/students/okasaki.pdf .
Very soothing and moves fast.
I think the premise was, before you had to actually do some work to create the thesis. And there was always the concern that yours could be one which was read deeply, so that thesis work had to at least show that you did some work. That there was meat behind the paper.
But now, it could simply be all a couple of prompts to an LLM.
The bar is just lower for not doing the work, now.
But really, that's the fact everywhere.
> AI is starting to show that the emperor called academia has no clothes
Why single out academia? We're seeing vast majority of knowledge work had no clothes.
>working your way through that system >pointless rituals >rethink their way of doing things, and they really don't like it
> So, it sounds like nothing of much value has been lost.
Do you actually mean this? To me this sounds like someone said "you never learn anything by reading a high school essay", and you reply "so stop writing them". The point is not the product, you obviously train people by making the product.
> AI is starting to show that the emperor called academia has no clothes.
Seriously, what are you talking about. Academia in the last century has been the most successful engine of knowledge and technology in human history. Lots of papers are junk, like lots of businesses are junk, lots of books are junk. But I don't know how any serious person can say academia has no clothes.
PhDs only were adopted universally in 1917 with some resistance and apprehension.
The issue here is that academia forgot what it was about a long time ago and is now having to face the consequences for a hundred years of bad decisions.
> PhDs only were adopted universally in 1917
In Oxford. That hardly counts as universal.
The largest output of a PhD has always been the training to the student, not the thesis itself, hence why we're called 'students'. Anyone claiming that a thesis can be generated by AI is missing the point. AI can also do everything an undergrad can do, we don't claim that undergrad education has been blown to bits. At best, we say we need better modes of evaluation, and perhaps that's true for PhDs as well.
I submitted my thesis this month at a QS top 10 uni after nearly 4 years of work. LLMs were available for most of that time. I don't really feel that it has diminished the value of my thesis by much really.
And how is the PhD student trained? By doing research and writing about it by him/herself.
> we don't claim that undergrad education has been blown to bits
But it has been. People pass CS classes without knowing how to write even a simple program. How do you think they'll fare?
As a meme said: you'd better start eating real healthy, because your future doctor will graduate using chatgpt.
>> we don't claim that undergrad education has been blown to bits > But it has been.
Let me fix it: pen and paper. There you go, bada-bing bada-bum. In our uni the exam for Algorithms and Data Structures (one that most people struggle with) is done on paper in an exam room. You better know your binary search.
People, especially wealthy, have been coasting through education paying someone to write their papers since Great Pumpkin knows how long. Now you can do same thing cheaply with LLM’s. The solution is pretty simple, let them fail. Have them show what they can do live in front of examiners. Here’s a computer without network card, only Python and SQLite installed with the Python basic documentation, build me X and prove it works in Z, Y, and Q.
I am studying my second degree and although there’s million and one ways to cheat, I won’t, since A) I want to learn this shot B) I am not sure LLM’s will be available with these capacity and with theses prices in the future, at least I won’t count on it.
Live exam, perfectly fine. You can use an LLM as support while you study, no problem, as a glorified search engine. On steroids. From the future. Then you prove you understood it.
Writing an essay, or a thesis, with LLM assist, is difficult to avoid in this scenario, though. It would require your prof, TA or examination committee to read the thing fully and understand the topic deeply. For 101 courses, that's doable, but for a PhD, that's so much work, that it'll break the system immediately.
Montaigne originally used essay (from French essayer, to try, attempt) to test hos on knowledge and understanding, his own capability to create new knowledge. In plain English, you test yourself if you can use your brain. That’s why it is so powerful to write a blog post about a topic you have just learned. Obviously, if you just generate the thing with LLM, there’s no point.
Essay is a fairly recent invention and perhaps it was not very good for testing students un the first place, since it kinda was not meant for that originally? So maybe we should test knowledge and understanding in different ways? In STEM-subjects this is not very hard, since basically you either can do the thing or not (like my example of coding above), but I have no idea how one would do this in history or philosophy. Should we go back to having long walks in the garden of the academy, where the professor guide’s you towards knowledge like some 21st Century Socrates? That doesn’t seem doable at this level of university students.
And how would you test PhD -level knowledge? You have to build a nuclear reactor from scratch in front of live audience? Or perhaps the oral examination have to become a real exam, not just a pointless ritual? Anyway, I think my point on using essay as a testing structure still stands. Maybe the examination should be like BJJ belt exam, where you have to show what you actually can do in front of peers and your teachers. A lot more people are likely to fail, but so be it. Maybe having a PhD stop being a joke then.
> we don't claim that undergrad education has been blown to bits
We don't??
You’re sort of saying that’s because your thesis had no value to begin with, because it’s mainly a teaching exercise. I’m not saying that it doesn’t have value by the way, I’m sure it was good, that’s just my read of your post.
It’s an interesting subject. Makes me want to vibe code a PhD generator just like in the tweet. Maybe I will.
They are saying that the value of most PhDs was in the learning the student got in the process, not in the PhD itself. Groundbreaking PhDs exist, but they are far, far, from the norm or expectation. Exceptional people doing exceptional work will always exist.
> your thesis had no value to begin with, because it’s mainly a teaching exercise
And that’s the problem in your understanding. Thinking that a teaching exercise has no value while OP was saying that IS the value. It’s missing the forest for the trees. The point of homework isn’t to solve the problems. I bet you the teacher assigning the homework already knows the answers. Just like the point of a marathon isn’t to travel 26 miles because you could just take a bus.
Well no, it’s a problem in your understanding of what I’ve written. The thesis and what’s learned along the way are separate objects with separate value, and you’ve decided to misread me in order to have something to be indignant about. However, to expand on a point I didn’t originally raise: learning itself is likely also reduced with AI assistance. That’s a fairly natural consequence of having to do less work yourself, we don’t retain information we don’t need to retain.
Pretty sure I understood you just fine as you’re just doubling down on your misunderstanding.
The value of the exercises you do in school (from sums to PhDs) is the learning you get from them. Good, because you need people to learn stuff, because the people that know stuff will be dead in a few years and you need to replace them.
Except, if machines can do the exercises up to the PhDs, you probably don't need those brains in the future, at least not as many as before. You might need only the brightest ones, and who knows for how long? The next ten years?
Yes, that at least is a valid argument. I do’t really agree with it though. Humans need meaning and purpose in life. Intellectual pursuit is that meaning or purpose for a lot of people. The Educational institutes might change drastically from how they have been in the last few centuries, but people will never seize learning. The act of learning is in itself incredible rewardig to many people.
> at least not as many as before. You might need only the brightest ones, and who knows for how long? The next ten years?
Then those with power will keep it and pass it on to their heirs, and some without will try to finesse it but the masses will be told “we don’t need you”. I think in a lot of ways, this is similar to how feudal societies were. Maybe humanity will pass through another phase of that. Maybe we already are. But I think it won’t last either.
I think they were saying the "largest" output was the training, which I agree with. But it's also the case that they nudge forwards thinking in various areas through the outputs of the research process, whether that's papers, talks or the thesis itself.
Socrates hated the written word . It weakens the memory. It does not talk back. Etc...etc... This is not a new problem folks.
> In the past, the path to a math Ph.D. cultivated resilience, resourcefulness, critical thinking, and a healthy skepticism (…) but suddenly you can produce a passable Ph.D. thesis with the push of a button
But actually nothing changed. Or maybe a path to "resilience, resourcefulness and critical thinking". Because human brains still needs to be shaped by years of training on some quality "literature", of some form.
We still must/want to human-[re]check important results, right ? And that require years of students time dedicated to memoizing facts and doing exercises in discovering already discovered results - learning and weights tuning, in the brains.
Yes, demotivator factor is very high or maybe just more visible then usual. Especially for brain paths forming - an that process is not quite stated in university and other education...
I think that when LLMs finish words and sentences shuffling and finds most of low hanging fruits in cutting edge of research ;) then only humans can move things forward, via abstractions, syntesis or old good paradigm abandoning. Hard to imagine LLM on their own "discover" something and then drops all that "literature" it was trained on as obsolote :) In next prompt it will happily return you old texts without any influence of just discovered paradigm shift.
In XIX century we got quite stagnation in science - it was belived that everything was already discovered, explained, just some few experiments are needed because some numbers do not adds up... And that proliferated to philophy and culture via some "proofs" for atheists. But in 1905 a paper was published... Too bad politicans do not get implications of that and still was pushing communism decades later...
So we realy want humans with brain pathways shaped mostly "old way" - the only one way available for human beings - by that training called "education". As always there is resistance and pain and attempts to find a shortcuts by cheating. Maybe this is time to clearly state that brain workings training is big part of education ? Just like in gym you are repeating to trying to lift weights up to your limit and even little above, with supervisor oversight.
> But in 1905 a paper was published... Too bad politicans do not get implications of that and still was pushing communism decades later...
The photoelectric effect/quanta? I'm not sure I understand the supposed connection to communism.
Einstein - definitive end of XIX ilusion and way of thinking that "We know everything. We are the lords of things, we own all of it! And other peoples too! We can engineer societes as we like it!". With strict oposition to old and humbler "God" or natural ways. With aggresive "we" that do not pay attention to rights of other we's.
And, IMO, Einstein, Godel and more discoveries should be reason to not continue communism stupidity. Especially to not continue - after year 1905 - social engineering plans of "world-wide GENOCIDE until everybody is good citizen".
Of course I'm talking about best-wanting-for-humans but still naive communism and not about a way to destabilize other countries by any means necessary. Such move did almost insta-karma effect on Germany: sponsoring commies in Russia resulted in nazi being sponsored in Germany...
Don't think someone want to deny societes, philosophers and politicians are drived by overall science view on reality. And what else could drive atheistic minds ? Not judging here, logic is everyones duty.
Also my "too bad politicians..." was obviously expression of a wish. Not only science influents politicians. But discovered facts (counterarguments in this case) about reality could be taken for consideration, right ?
> some quality "literature", of some form
In my opinion, we will continue to need fully educated people who read books and peer-reviewed articles. We need literature, not "'literature', of some form."
I have a PhD and I completely support AI disrupting the field.
If AI disrupts your field, the culprit is most likely not AI.
“If algorithmic targeting disrupts your society, the culprit is most likely not algorithmic targeting”
Or replace AI/algorithmic targeting with tech in general and the disruption target with whatever it targets and you’d realize the problem with the sentence.
Technological advances have disrupted plenty of fields. That doesn’t mean those fields were fundamentally flawed. Every arena has a certain degree of dysfunction. AI has its own massive share of issues already. But that doesn’t negate the whole field.
Take the classic example of the Travel Agent. They are all but extinct because of technology. Yet they did serve a legitimate purpose before. Yes, plenty of them were middlemen who didn’t care, but also plenty were passionate about organizing travel plans and helping people arrange their travels and vacations. Plenty of people using AI today are also middlemen between you and Claude who also don’t care
If tech is disrupting X, that’s a problem with X” tends to be said by people who have X comfortably sorted themselves (e.g., already have a PhD that counts as it’s from the pre-LLM days).
In the end it depends on whether one considers technological progress or human flourishing to be the end goal. Surprise—they are not always aligned, and sometimes are at odds with each other.
Agreed. It can also be people who have problems with X for other reasons and like to see X “disrupted” (e.g. a disgruntled grad student who got screwed over by a bad advisor or a hard working and capable office employee who got passed over for a less capable, less hardworking PhD holder).
Like I had many bad experiences in cabs in my younger years. It was just bad luck really, but I hated cabs for the longest time because of that and completely dreaded needing them. When Uber was first coming on the scene and cab drivers were protesting against it, I did think to my self something along the lines of “if an app can destroy your business, the problem was in your business” but I really just didn’t like cabs for other reasons. And the irony is now Uber is the same. When I land in my airport, I see crazy lines for rideshare with people waiting on cars to come pick them up, while there is a cab on the other side that is ~$15 cheaper somehow and ready to go with no lines. And I get to tip the driver those saved 15 bucks
Travel agents are not extinct AT ALL. They pivoted from selling margin-less flights to design more curated travel packages. Different and nichier market for sure, but not extinct.
Fair point. They also still exist for corporate and business travels. But it’s fair to say the field did get massively disrupted (god I hate that word) because of technology.
Could you expand on that?
In no way getting a PhD or doing published research is an act of science or human progress, and it's been like that for probably 20 years.
It's an act of coordinated methodology in showing respect to previous peers by acknowledging that you read what they wrote when they were acknowledging even more previous peers.
Every attempt to do something new is rejected unless you take 99% of something that has been done and try to add your 1% to it.
But the thing is that you don't even want to do that, you need to either have a publishable/defensible thesis, or if you already have a PhD, pursue a tenurable track and grants which never collides to productive human progress.
So yes, AI could very well run tenure track career more efficiently and with better results. It's not AI the problem: it is the checks and performance indicators that are in place that make it very easy for AI to dominate and very exhausting for a human to follow.
A good researcher with good AI knowledge would (and should) dominate their field.
* Medicine is luckily saved from this, with some exceptions.
Maybe old methodologies will be discarded and new ones will emerge.
Thinking back, when I first started teaching myself programming, I didn't know what to learn, so I explored the history of programming and organized it as I went.
One of the most striking things I remember is that when Stack Overflow first launched, quite a few people opposed it.
Also, I recall that in ancient Greece, Socrates criticized writing, saying it would weaken human memory.
When SO first appeared, there were many who insisted that the only proper programmer's way was to RTFM, deeply understand the system's fundamentals, and then write code. I wasn't from that generation—I belonged to the copy-paste-from-SO generation—so I can't say for sure, but I found it quite fascinating.
The cost of that friction could only be borne by a very small minority, and that minority could guarantee quality. That's why scholarship was something only the elite could pursue—and to some extent, it still is.
In the past, tasks were painful and high-friction. The results were filtered through that process, accessible only to the few who could endure it. But we tend to mistake those inefficient drops of sweat for quality. In reality, just as writing didn't diminish philosophy but rather created systems like law and philosophy, this might just be another turning point
I think Mr. Lemire's post is similar in spirit. In other words, when friction increases, the cost of production for producers also rises. So back then, everything produced through that high-friction process was easier to quality-control. But that's no longer the case.
And actually, universities were originally about 'holistic education,' but these days, they've become more about training talent for industry and managing human resources for the job market. That shift has caused problems.
In that sense, it's only natural that these problems arise at the intersection of academia and industry.
Industry usually demands 'people and technologies that can boost productivity right now.' Meanwhile, academia should ideally pursue problems worth exploring over the long term, even if they have no immediate utility. But the current state is a product of compromise.
Once university evaluations, student recruitment, research funding, and employment rates become tightly linked to industry demand, the latter starts to pressure the former. And under those conditions, the current outcome is almost inevitable—because industry increasingly wants to churn out degree stickers at lower and lower costs.
In the end, a different methodology will be needed, and whoever proposes it will become the game changer. Then new schools of thought and methodologies will emerge based on that person, and they'll gain enormous fame. I'm curious who that will be.
New things are always born by laying the past to rest. I'm always waiting for that new methodology.
>Industry usually demands 'people and technologies that can boost productivity right now.'
Industry demands a giant sorting machine and that is what they got. If we acknowledged and accepted this we could come up with a much better solution than what we have.
"...and that's a good thing."
[dead]