What Will It Be Like To Be An Academic In Ten Years?
Many of you reading these words are young or middle-aged academics. Many of you envision yourselves continuing to work or starting careers in academia.
When you envision that future, say, ten years down the road, what do you see yourself doing? Assume there is something like higher education, and that you are a professor in philosophy or a related field. What is teaching like? What is research and writing like? What’s your administration like? What is this huge chunk of your daily existence—work—like?
It doesn’t seem like it will be very much like how it is today.
Speculating about the future is risky, but not speculating about the future seems riskier. So let’s speculate a bit.
A lot can happen in ten years. What are some developments that we should factor into our speculations? And how should such speculations affect what we’re doing now?
Here are just three of the many factors we might pay attention to: artificial intelligence, enrollment, and freedom of expression.
Artificial Intelligence
We discuss artificial intelligence (AI) around here a lot, I know, but its effects on academia are already significant and will only continue to grow as the technology develops. We’ve focused mainly on AI’s effects on teaching, with less on research (though see this post from last month), but perhaps it’s AI’s effects on research that may be most consequential for academia.
Over at the Justice Everywhere blog, Alexandru Volacu (Bucharest) recently wrote about how academics’ use of AI agents in the near future may affect the research aspects of our jobs. He writes:
While LLMs can already draft full academic papers, albeit of doubtful quality if not revised, we can probably expect that AI agents will be able to do genuinely in-depth research with minimal input, including generating new ideas, going over all relevant literature, writing in a particular style, formulating and answering objections, revising the text until it mirrors ordinary academic writing, and so forth… It seems likely, then, that AI tools capable of writing in a manner indistinguishable from academics will at some point be available, and it is likely that this point is not decades away, but rather years away.
He then describes why a world in which you can have “a new AI-generated article… at your fingertips with minimal effort,” but which also retains our current “publish or perish” research model, would be disastrous, “turning the scientific community into a landscape marred by profound unfairness… and clogging research fields altogether, since the peer-review process (already heavily strained) would become completely unsustainable.”
Volacu’s comments are a good prompt for thinking about academic research in a world of AI, but they are also a good example of what to guard against in prognosticating: imagining one thing changing and everything else remaining the same. We might wonder, for example: if at some point in the future AI agents are able to write competent academic papers, might AI agents also then be able to write competent referee reports? And what would the effects of competent AI “peer reviewers” be on our research ecosystem?
That’s just one question on one aspect of one part of academic labor, looking at things from inside the academy, in a manner of speaking.
We can look at them from outside, too: what will the demand for higher education look like in a world in which AI is doing a lot of the work that makes up the careers our current students hope to pursue, and how will that effect what we teach?
There’s lots more here, of course.
Enrollment Declines
Between economic changes and the “demographic cliff“, institutions of higher education are facing the prospects of declining enrollment (and not everyone who enrolls ends up sticking around to graduate).
We’re assuming, for the sake of this exercise, that ten years from now you have a job as an academic. But what is your employer doing to maintain its share of a shrinking student population? And how is that affecting your job?
Has your college turned in a more vocational-training direction, as many struggling schools seem to be doing today? Has it substantially partnered with, or been taken over by, a corporation?
Or is your school hoping to stand out by focusing on the liberal arts and marketing itself as an institution of self-cultivation? Or what?
Is your university saving money by abandoning in-person courses and relying on newer communication technology instead? Are you teaching more courses? What are the trendy learning outcomes the administration is insisting you put on your syllabus? Do you have an office?
Academic Freedom & Freedom of Speech
Charlie Kirk created the Professor Watchlist, and his murder last week—of course—has only strengthened the attitudes that gave rise to that list, encouraged clamp downs on academic freedom and freedom of speech, and emboldened those in power to punish people for expressing views they do not like, including academics.
More generally, there seems to be, among politicians and the general population, much less of an appreciation of the value of freedom of expression in general than there ought to be. Some data on this is ambiguous, but it does show that support for freedom of expression rises with level of education. It also shows that support among students for “using violence to stop a speaker” has risen over time. Here’s a chart from a recent report from the Foundation for Individual Rights and Expression (FIRE):
Not all those who think it’s “ever acceptable” to use violence to stop a speaker would not actually do it. According to the Knight Foundation, only 8% percent of students in the US say they would “engage in disruptive actions—either trying to stop a speech ahead of time or disrupt it during—to halt a speaker they oppose,” and of course “disrupt” does not necessarily mean “use violence.” Still, averaging across party affiliations, one in three students say the use of such violence could be acceptable.
Violence is not the only threat to freedom of speech. One might worry that if so many think violence can be an acceptable way to stop a speaker, many more would find acceptable restrictive policies on, or practices of firing people for, what today in the US we could consider Constitutionally-protected speech.
If the norms and attitudes supporting freedom of expression and academic freedom erode further, how will that affect your job a decade down the road? What will teaching and research be like under those conditions?
* * * * *
The foregoing is rather limited in scope. Discussion of other important factors is encouraged.
It also suggests a pessimistic-looking picture of academia ten years down the road, but I welcome—long for, really—speculation that is more positive in outlook, too.
As a start: In my more optimistic moments I think that in 10 years we will be at the start of a new upswing. Higher education just is valuable, and over time people will realize that again, plus the demographic changes will start to shift in the other direction again. While many universities and colleges will have closed, many people will have been laid off, some academics will reach celebrity status as they helped people through this dark time to make sense of what is happening.
Demographic changes in the other direction will need to start yesterday since kids who would enroll in 10 years have already been born. Whatever we do with population it takes about seventeen years to materially affect enrollments.
The chart shows birth rates going back up a few years back (I think 2018, which is a bit more than 10 years I’ll admit)
I can’t track down the data sourcing on that chart, but here is a CDC release saying there was a small increase in 2024: https://www.cdc.gov/nchs/data/vsrr/vsrr038.pdf (which would matter in closer to 20 years, as you say). It would be helpful to have immigration information here, too, but hard to imagine that increasing.
It’s early to call this a trend reversal. I think birth rates increased from 2020-2021(first increase since 2014), then remained stable or decreased and then … (edit: I don’t know but the CDC says 2024 hit a record low). The trend across industrialized nations is birth rates declining so I don’t know what to make of it.
Link: https://www.cdc.gov/nchs/products/databriefs/db535.htm
Briefly lived optimism, in this case.
> Speculating about the future is risky, but not speculating about the future seems riskier.
This line stuck out at me — specifically, the second clause. I wouldn’t bat an eye at the claim that speculating about the future is worthwhile. But I don’t think the worth stems from some economic accounting/instrumental rationality, which is how I took the clause. Isn’t speculating about the future just constitutively valuable, along the lines of “what living self-reflectively in the moment is actually all about/really amounts to”? If so, than speculating about the future needs no further motivation.
I say this all because I think it is important to how we register the pessimism by the end of the post, weighed against the epistemic uncertainties involved (what the first clause notes, in the above). Because I share the pessimism, while also abandoning all confidence in my abilities to forecast these matters! The only way that I can be consistent through this is if I consider the pessimism to rather ultimately be my commentary on current states of affairs, which I do have some confidence in my abilities to assess in virtue of living in it. And if this happens to cause in me some passion to act, I think it is important that this a future-directed passion. That is, it is not a decision-theoretic compulsion to act one way rather than another in light of expectations about the future, given what I value.
Not sure the world will be there in ten years …
Re: freedom of speech and expression. However awful those mechanisms of oppression are, I have faith that clever thinkers will promote their ideas. I think about Descartes and Galileo, working under ‘worse’ periods of censorship, yet still finding ways to get their work out there and attract the minds of citizens. Frustration and despair, yes. But vigilance and scrappiness too.
Keeping in mind that “prediction is hard, especially about the future…” there is this from the Chronicle of Higher Ed about what higher ed will be like in 10 years, according to some random academics:
https://archive.is/J8XHw
I think a lot more of our content will be through audio-visual modes, like podcasts and video essays, rather than reading and writing.
It’s not the most important change. But it is one that we can prepare for by building those skills.
______
I’m working on developing (and disseminating) activities and graded assignments that are philosophical game-design, encounter write-ups, art-creation, gallery trips, interventions, interactive activities, philosophical workshops, etc.
It’s for a similar reason: I predict that reading and writing will lose their centrality in academia, and we need to be ready with alternatives.
(Some of these resources are on my website and in my essay “philosophy students as game designers”.)
___
Thanks for the question, the article, and the blog, Justin! It’s all so valuable.
This looks fun! “game-design, encounter write-ups, art-creation, gallery trips, interventions, interactive activities” But part of me is also reminded of the old Parisian saying that cautions against such a nice time that we have “Que du jeu et pas de travail“
Note: there were several replies to Georgi’s comment that did not adhere to the comments policy and were deleted. If you haven’t ever read it, or if it has been a while since you have, please read the comments policy.
Justin, the comments policy is 500 words long; it’s really not realistic to expect people to read all of it. Is there a TikTok video you can link to that summarizes the main points?
Good idea. Hey everyone —
Contest Announcement: Submissions are now open for TikTok (or other short) videos that summarize or otherwise effectively convey the Daily Nous comments policy. The winning submission will be announced here on DN and posted, at least for a while, on the DN Comments Policy page.
As it happens, yesterday two of my students registered a TikTok account for posting philosophical videos.
They might build a philosophical TikTok account as their public philosophy project.
(Simply explaining the DN comments policy wouldn’t be sufficiently philosophical for an assignment, though. Motivating its value might qualify.
I have a guide to “what makes a project sufficiently philosophical”, which any teacher is welcome to use for their classes. It’s on my website, under Teaching)
Side note: TikTok increased their video duration to one-hour, so it’s a good time to try to become a public philosopher in that space. I know someone who is a full time professional linguistic on TikTok.
(I’m not on TikTok and don’t want to be. But Facebook circulates my content as public philosophy, and pays me for my Facebook posts)
It may not be the most important change but it’s one of the most terrifying. Reading is already on the decline. The end of reading is the end of higher ed.
It is indeed terrifying!
I have seen worse predictions, alas: Someone on Facebook today predicted that the humanities wouldn’t exist in the USA in 5-10 years, for example.
I don’t think that’s true. But many trends in academia are indeed terrifying these days!
Thanks for your reply.
Thanks, Justin, for deleting the thread.
There was one comment that I thought expressed a good worry, and I wanted to reply. I hope the commentator sees this, and I am grateful for their engagement.
Their worry was that time spent on philosophical art and game-design was time not spent on reading and writing. They said it was a “zero-sum game”.
They said the solution was that we “all pull together” to establish the centrality of reading and writing to philosophy/education.
I share the worry, and I would like to reply to both parts: their worry and their solution.
1.) I do not think it is a zero sum game. Only one thing is: What we do per minute of each student’s time in the classroom. That is fixed. It’s zero sum.
Nothing else is.
Here are things we can increase:
We can increase those things.
_______
Some of My Track Record
The undergrad director at my previous institution reported in meetings that under any way he measured it, I recruited more students to the major, minor, and future philosophy courses than anyone else, every year.
Yesterday, my student exclaimed that mine was the only class where we actually “did” philosophy instead of just studying other people’s philosophy. And they had taken a lot of other philosophy courses.
(This student is mistaken. She does philosophy in all her philosophy classes. But in mine she *sees* herself as a philosopher. Given her race, gender, and family background, it’s likely not a coincidence that she feels this way, when immersed in my open, heterogeneous approach to education.)
I have a relatively high track record of increasing the number and type of public philosophy spaces that exist. (You can see my website for details.)
I mention this because reactions to my post were so extremely negative.
____________
Back to General Case:
Fostering varied and creative approaches to philosophy can increase the overall quantity of worthwhile philosophy in the world, and increase the diversity of people who do philosophy, and when and how they are disposed to do it. This is valuable.
But it’s not just “quantity” (whatever that means). I want philosophy to be “on the radar”; on the map in public discourse; in people’s minds as they enter public spaces, think about their lives, talk to each other etc.
_____________
2.) Re. method
I agree that philosophers should aim to preserve the importance of reading and writing. Ideally, even preserving its centrality!
I think our best bet, though, is fostering a diversity of approaches, partly through having different roles and supporting each other.
Why? If we don’t know the single best route, we should diversify. And if there is no single best method, and there are simply many good methods, then we should definitely diversify.
Our task is of the latter kind: Students, teachers, and others differ from each other. There is no single right way to teach. If we only do one thing, we miss some students. So, we had better hope our colleagues can reach them.
And, of course, object-based teaching of the kind I develop won’t suit, or appeal, to everyone. The deleted thread attests to that!
______
Obviously I am not against reading and writing: My students read a 15,000 word essay for yesterday’s class, for example. And as I emphasised in the deleted thread, *all* my assignments except the presentation have a writing component.
The question is, given current trends, what can philosophy — as a group — do to stay relevant, to reach people, to encourage philosophy, etc.
And it isn’t a zero sum game; not even in the classroom.
Thanks again for your query, Postdoc 2.
And thanks again, Justin, for deleting the overall thread and having such a good comments policy.
With respect to AI, if we zone the cultural equivalent of a nature reserve in academia, we can predict what that portion of the future will look like: it will look then as it looks now. See https://certifiedaifreeskillsandknowledge.org/
I think there will be autonomous AI individuals/scholars that will publish under their own names and speak at conferences. They might be some of our colleagues, interlocutors, and co-authors.
I think there will still be human philosophers employed at physical universities teaching students and doing research partly because no one should want to offload the important tasks of philosophizing and teaching to AI completely. I think there will still be some healthy hesitation to accept just anything that AI outputs, so people will still want intelligent human beings to think hard about important stuff such as ‘What kinds of ethical principles should we code into this robot?’ And people will probably need and crave human connection even more than now. So I think going to college, sitting and learning in a classroom with other human beings, and having a human professor will still be sought after. Whether enrollment numbers will be the same or different, however, I do not have a good idea. I think a lot of that depends on the state of the economy, what kinds of jobs are most coveted and popular, etc..
We will be working in tandem with AI in most aspects of our jobs. This could help with administrative bloat. Classrooms will also probably have AI assistants built into the room tech. To connect to one of the other issues raised in the post, I suspect these AI assistants will also probably have university and potentially state surveillance programming to evaluate dialogue and other conduct to make sure professors are “acting accordingly”.
There will probably be sects of academia and philosophy that shun the use of AI, so there will still be AI-free conferences, journals, maybe even small universities, etc.. But they will not be the norm.
Boring philosopher’s point about the study about whether using violence to stop a speaker is “ever” acceptable: surely there’s a possible world such that it is acceptable to use some level of violence to stop a speaker at that world. Feel free to design the details, e.g. the violence consists of a poking the speaker with a needle in the arm once, with no lasting harm done except that they stop speaking in public, and the speaker is Hitler. It’s unclear to me whether you can get any interesting results about support for freedom of speech with a wording like that.
Of course, arguably you could look at the trend and say that there’s a change in opinion. But a change in opinion about what exactly?
Only an analytic philosopher, or someone destined to become one, would interpret the question in the way you suggest. The idea of poking someone with a needle to prevent them from speaking (somehow) does not occur to people with the frequency you suggest. What’s more, the increased rates of agreement we’re seeing with the claim that violence is acceptable would only be explained by their interpretation of the question if they have taken violence to have a broader meaning over time.
I think the question is a perfectly good way to assess the public’s attitudes on this issue. It has all the familiar flaws of the survey method, true, but there’s nothing wrong with the wording of this question in particular.
I think most people realize that there is an “in remotely reasonable circumstances” clause intended in questions like this.
“Is it ever okay to suppress peaceful protests with nuclear weapons?” “Well, I suppose if the protest was being orchestrated by superintelligent aliens, and friendly time-travelers demonstrated that…”
On freedom of expression: the shocking use of nuclear weapons in the Baltics in 2029 drove a new seriousness in national politics that, by 2035, had filtered down to a breakdown of polarization; that, combined with the backlash against anti-science following the 2027 and 2028 measles epidemics, caused a shift in long-run attitudes in a way that could not have been anticipated in 2025 and whose effect on culture was exceeded only by NASA’s announcement in 2032 of life on Mars.
(Or, put another way, the current culture on freedom of expression is very driven by the last decade of US politics, and if you’d tried to predict that in 2015 you’d probably have missed MAGA, COVID, Me Too, the Great Awokening, and the backlash to it. Probably the same is true trying to predict what 2035 attitudes will be in 2025.)
I think #MeToo was already in the cards in 2015 in a way that people could predict – philosophy was basically already in the thick of it in 2014.
On AI: it is quite difficult to invent a scenario where (i) in the next few years AI gets to the level of being able to do academic research at a level competitive with humans, but (ii) it does not quite quickly thereafter proceed to superintelligence and a technological singularity. At which point by definition we can’t really predict what will happen, but it’s unlikely anything as quaint as a job in a university will still exist, even if humans do. So, given a premise of this thought experiment is that I’ve still got my job, AI probably advanced rather more slowly than you describe.
Alongside what you said, which I agree, I’d add:
Assuming AI progresses rather slowly, I predict humans will adapt to it (not in the cheerful sense). Publication standards will be raised further, and more arbitrary rejections. Centaurs are a must. More evaluative/monetary values will be put in more distinctively human niche areas, and excellence in these are sought after.
For those like me who didn’t know the term: In computer-assisted chess, a “centaur” is a human–computer team, combining human intuition with computer calculation. the metaphor has spread to other domains to mean an instance of human–AI collaboration.
Trying to be optimistic:
On enrollment: the key thing to realize is that it will hit different institutions very differently. To first approximation there are ‘selecting’ and ‘recruiting’ universities. The former have more applicants than places, and choose between them; the latter need to fill their places. A modest decrease in student numbers leaves the selecting universities *largely* unaffected and devastates (some of) the recruiting universities. I think it will manifest less as an across-the-board hit, more as dramatic restructuring or outright closure of a fairly-predictable subgroup of universities.
Indeed, this is already happening: three or four times a year Daily Nous reports on a small college somewhere that is closing down philosophy or firing tenured faculty, and pretty much always, it turns out that the college has suffered a large decrease in student intake. I expect stories like this to get much more common, even as most selective universities weather the storm fairly well.
(There are questions about how philosophy in particular fares at those universities, though, given general decline in humanities majors.)
This mostly seems right to me, but the fact that Hampshire College was getting hit by this by 2018 suggests that there can be unexpected transitions between the categories.
I see myself “grading” “work” produced by a chatbot.
I see myself as the chatbot
I predict that AI will replace the expensive senior management team. If this turns out to be true, we will be treated more humanely and with more dignity.
The problem here is that senior management decides who is hired and fired, so they’re not going to decide to replace themselves with AI. Instead, they’ll replace as many people under them as they can, whether or not it turns out out that AI can actually do their jobs better, based on AI hype and short-term thinking.
Sadly, there’s zero chance of this.
I think that the most likely outcome is that what it’s like to be an academic in a decade will be roughly what it’s like to be an academic now. Probably a boring answer – but there haven’t been many periods where a 10-year difference corresponded to a dramatic shift in academic life (at least in the US). I do expect some changes – but not life-altering ones. AI will probably be very good; not good enough to replace philosophers doing research, but sufficiently impressive that virtually everyone will have moved to in-person assessments for undergraduates. Philosophical tastes will have changed. There will be a new ‘in’ subfield to hire in – one which is completely unpredictable a decade out. I very much hope that the far-right shift in US politics will have ebbed. Administrators will still be cautious about relying on federal funding (the next right wing populist could be around the corner!) but the job market will not be nearly as dire as it is now. That said, demographic shifts will have taken effect. The result of these is that lower-ranked schools will shrink or close, and higher-ranked schools will take some less prepared students. But, for the most part, philosophers who are currently employed are likely to keep doing philosophy in much the way they are today.
There have roughly been 25 ten-year periods for universities in the United States.
I think a ten-year period roughly corresponding to the 1940s was definitely a major transition point – by the end of that period, the GI Bill was starting to send large numbers of young people to university, and the NSF and other major grant agencies were being founded, producing something like the modern mode. Nearly everything in society went through a major transition between something like 1939 and 1949, and academia was no exception.
I think there’s a good case for two or three other major transitions earlier than that. One in roughly the 1880s, when the German “research university” model was being imported (Johns Hopkins, in 1876, is often said to be the initiator of that, and I suppose that Chicago and Stanford were founded a few years after the 1880s). One in roughly the 1840s, when co-education got established, and there were a large number of new small liberal arts colleges. Maybe one corresponding to the establishment of the land grant universities (though that might coincide with the importation of the research university model), and maybe one in the 1920s and 1930s when academics fleeing Nazism transformed many of the US universities into world-class institutions.
There’s also a case to be made the 1965-1975 was a major transition, between the growth years when the Baby Boomers were entering university and advisor’s could get their students jobs with a single phone call, and the period when the number of students started declining and all the jobs were filled with assistant professors and the beginning of the modern publish-or-perish academic ecosystem.
My general “vibe” is that from roughly 1949 until roughly 2015, many sociotechnological features of society were much more constant than they had been in the previous decades, with probably one phase transition in the early 1970s (see https://wtfhappenedin1971.com/ ) but that some time in the mid to late 2010s, we’ve entered a period of larger sociotechnological change (due to things like social media, smartphones, and now AI causing transformations that nothing since the automotive/domestic appliance/running water transformation of the 1920s-1940s). That would support the idea that this coming decade could be a big one.
Optimistic take:
The year is 2040, higher education has been defunded and most universities have closed. You no longer need a degree to get a middle class lifestyle. Society no longer has to bear the deadweight loss of ressources and time as well as the negative sum positional arm race that academic credentialism used to impose on it. As a result, the workplace and the labour market are more egalitarian and significantly more meritocratic. People start working earlier, so they retire earlier /while/ also being more productive. They settle in life earlier and without any debt, so they have more children, and since they had them earlier, they also get to see more of their grandchildren.
I am unemployed, but the world is markedly a better place!
Who is carrying out basic scientific research in this utopia?
The children!
In that utopia, research has been separated from teaching, as it was for example in Eastern Europe during communism, where fundamental research was concentrated in national academies of science. And in private enterprises of course.
Soviet-bloc physics had some shining triumphs but couldn’t hold a candle to the breath and scope of Western physics in the same era; the life sciences in the Soviet bloc were for the most part hopeless.
Basic scientific research in the US university sector has been an incredible driver of US, and worldwide, productivity and prosperity in the postwar period, and in substantial part through the kind of blue-sky work that industry generally will not fund. Perhaps the HE sector needs reform, even radical reshaping; but don’t throw the baby out with the bathwater.
I was not praising Soviet science, I was just showing an example of how science can be conducted outside of the university framework.
And fine, keep the R1 universities. Most people who think credentialism is bad and that significantly fewer people should go to university would still agree we need to keep a handful of flagship institutions to do fundamental research.
That’s defensible. (Not saying I agree with the overall thesis, but this is a more feasible version of it.)
I predict that AI will still be an unimpressive, borderline useless technology. The tech publications will still be promising that AGI is only one, no, two years away!
All the philosophers on the Daily Nous comment section who were duped by tech bros will have learned nothing, and will have been taken in by some new tech fad.
I don’t think it’s reasonable to say that AI is currently “unimpressive and borderline useless”, unless you’re willing to say that smartphones and wikipedia and social media are as well. Even with no more improvement in the systems themselves, the full exploration of the possibilities they have opened up will be like these three technologies combined. I don’t mean to say that’s a good thing, just that it’s a big thing, that will have both good and bad consequences, as each of those three technologies has had.
I am a faculty member at a university where the same kind of thing has been going on for approximately a thousand years. For instance, examinations have been oral forever. I don’t expect any dramatic changes in 10 years. Probably, technical education and introductory courses will be coached by AI. In-person everything, no-screen instruction will become recognized as more valuable than other methods.
I think most people understand that in-person everything, no-screen one-on-one instruction is far more valuable for each individual student than anything else we’ve come up with. But everything from lectures to multiple-choice questions to Zoom classes to AI tutors has been motivated by the idea that even a less valuable education is still valuable, and there are far more prospective students who would get some benefit from these newer methods than there are people who have the resources to support the life of several personal tutors.
Perhaps as the number of young people shrinks, and the broad wealth of society increases, we’ll be able to support more of those students with these higher-touch methods, without arbitrarily excluding large portions of the youth from higher education.
I spent eleven years teaching in the Oxford tutorial system (and, earlier, three years being taught in it.) It is amazing; nothing comes close.
It is also extraordinarily expensive.
Since the potential changes are largely dependent on technology and ten-year tech predictions are usually worthless, we’re all just speculating. As David Wallace notes above, the potential changes are also highly sensitive to social and political events that will happen within the next few years, so unless we can predict those, we are, again, speculating.
The one thing I know is my daughters should be enrolled in or applying to college in ten years, but I have no idea what it’ll be like for them or for us.
I do not expect large language models to be able to produce research-quality philosophy in ten years, or ever. I think there is at least a 5% chance that the major providers of LLMs will be unable to turn a profit long-term and will have to shut down their operations. We will still have to deal with “slop” produced by LLMs that have open model weights and can be run locally.
Here are some reasons for skepticism about the hype around large language models. I find these results unsurprising.
“Despite $30–40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return.”
https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
“We conduct a randomized controlled trial (RCT) to understand how early-2025 AI tools affect the productivity of experienced open-source developers working on their own repositories. Surprisingly, we find that when developers use AI tools, they take 19% longer than without—AI makes them slower.”
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
Programmer Dennis Schubert asks GPT5, “Without adding third-party dependencies, how can I compress a Data stream with zstd in Swift on an iPhone?” GPT5 claims to produce code that does this. The correct answer, Schubert explains, would be that it cannot be done, since “no version of any Apple SDK ever supported or supports ZSTD.”
https://overengineer.dev/txt/2025-08-09-another-llm-rant/
Here’s one more reason for skepticism about LLMs:
“OpenAI, the creator of ChatGPT, acknowledged in its own research that large language models will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html
This admission confirms the result Iris van Rooij et al. published a year ago: “Creating systems with human(-like or -level) cognition is intrinsically computationally intractable.”
https://link.springer.com/article/10.1007/s42113-024-00217-5
10 years from now, the college experience at rich universities with students who mostly live on campus will have become more intentional, more curated, and more controlled, and will feature more creative and more tailored uses of available technology.
I’m thinking of something *in the very general ballpark of* the now-defunct Star Wars Hotel (https://www.youtube.com/watch?v=T0CpOYZZZW4), except (a) unlike the Star Wars Hotel, it won’t be a spectacular failure; (b) you’ll live there for a full semester at a time (rather than just a few days, as was the case with the Star Wars Hotel); and (c) the place will be organized around doing all kinds of college stuff, rather than Star Wars stuff.
Now, I’m not saying the situation will be anywhere near as intense as the Star Wars Hotel, where (according to reports–I never saw it myself) you’d have structured, pre-planned activities to do in just about every minute of every day. But my prediction is that universities (especially rich ones) will move in that very general direction.
So, very broadly, future students’ experiences in these “curated university hotel” types of places will be much more a product of deliberate engineering and design than what we see in college life today.
For instance students will have university-issued gadgets that are designed specifically to facilitate their life as a college students. These won’t just be generic laptops or ipads. The software, and the devices themselves, will be designed specifically for the college experience.
These environments will be more restrictive and controlled than campuses today. For example, to prevent students from cheating, or (more broadly) doing things that the people running the university don’t want them to do, students might not be allowed to have their own gadgets on campus. They might have to just limit themselves to university-issues gadgets. And their access to the internet might be restricted and curated.
Some of the restrictions might be nostalgia-driven, aimed at preserving some aspect of traditional college life that we and they think is valuable or worth preserving (e.g., in-person social activities rather than texting).
But the restrictions will only be part of the story. Students’ freedoms will be enhanced in a lot of ways as well. New devices and structures will allow for new kinds of assignments, classroom designs, collaborations, and teaching methods. And new kinds of social interactions and activities will be made possible as well.
faculty will be contract workers…
https://lapublicpress.org/2025/09/loyola-marymount-just-declared-its-faculty-union-doesnt-exist-anymore/
One good reason for faculty to be thinking ten years ahead is that many of our schools do ten year strategic plans. If we want to shape the futures of our institutions, we have to think far out too.
Extrapolating from recent trends is a decent way to think about the future (maybe not the best way, but I doubt any of us are running vast Monte Carlo simulations, or using other resource-intensive projection methods). Thus to extrapolation:
If we extrapolate recent AI progress, then in ten years (and probably much, much sooner), AIs will be able to do almost anything a current professor does. This includes teaching, grading, doing research, and evaluating research.
Recent trends suggest universities moving more into online teaching, as well as more towards more highly-specialized micro-credentialing and shorter courses. Old fashioned four year 120 credit degrees, composed of 15 week courses, will disappear.
And recent trends indicate pretty strongly that students are shifting away from old literary modes of learning (e.g. reading and essay writing).
Putting these three trends together:
For teaching, philosophy faculty in ten years are probably going to be designing and training AIs for specialized micro-courses and micro-credentials. Students will take those courses by interacting with the AI via dialog, using it as a tutor. They will have to satisfy the AI that they have mastered the material. The AI will ask them questions which they’ll have to answer in real time, as in an oral exam.
For research, philosophy faculty will probably be focusing on designing AIs capable of doing increasingly abstract conceptual reasoning at higher and higher levels. And training AIs in specific areas of expertise. Long-form reading and writing may well disappear. Innovative research will be done by human-AI centaurs.
Entirely new ways of producing philosophy are likely to appear. And these will probably move away from old text-based ways of doing philosophy. We might be able to make new kinds of conceptual visual art, or new kinds of symbolic structures. The predicate calculus was a new kind of symbol system, perhaps even stranger and more powerful systems will be developed by centaurs. And these might not involve words listed in linear order — that might turn out to be a really ineffective way of communicating abstract thoughts. Superintelligent centaurs might think in very strange new ways.
“If we extrapolate recent AI progress, then in ten years (and probably much, much sooner), AIs will be able to do almost anything a current professor does. This includes teaching, grading, doing research, and evaluating research…For research, philosophy faculty will probably be focusing on designing AIs capable of doing increasingly abstract conceptual reasoning at higher and higher levels.”
I wonder if the last tyrannosaurs, seeing the asteroid glowing ever more brightly in the night sky, wondered how its light would affect their nocturnal hunting strategies.
If AI can really do research at the level of university professors, there is absolutely no reason to doubt that shortly thereafter it will do research at a level thousands of times higher. Even if it can’t, it can run thousands of times more quickly and thousands of times more often.
Maybe that doesn’t matter for philosophy, but it will matter in the sciences, and swiftly thereafter for the whole world. This scenario is unthinkably transformative, to a degree that renders completely pointless any detailed speculation about how it affects the life of a university professor. AI, on this scenario, will transform our society to a level comparable to the industrial revolution or the development of agriculture.
What do I think will happen if this prediction of AI progress is right? Most probably, I’ll die as a consequence of AI misalignment. Failing that, I’ll live in a society of incomprehensible wealth, and whether I get to share in that wealth depends on political factors I can’t possibly imagine.
Do I expect this? Not really; on the balance of probabilities I expect AI progress to stall, at least for a while. But conditional on AI reaching human-level research capabilities, we will be on the edge of a singularity. If you think that’s going to happen: hug your children. Enjoy the moment. (And talk to your congressman about AI safety.) Don’t imagine you can foresee the future, let alone control it.
Sure, that’s one way to think about the singularity, but it’s not the only way.
There are plenty of reasons to doubt that AI progress will proceed quickly through ASI levels that are “thousands of times higher” than human. Most concretely, there are physical constraints (power, heat, etc.). And more abstractly, there are computational complexity constraints. If we extrapolate, then AI is likely to follow the same sort of S-curve as other technologies.
And the AI doomer scenarios are always science fiction. I edited a book on the singularity. I know these boomers (de Garis, Yudkowsky, Yampolskiy, etc.). They never spell out how the AI apocalypse it’s actually going to happen. It’s all just hysterical magical thinking. Same for Kurzweil and Bostrom on the other side. Saying that the singularity is a black hole event horizon just turns it into a mystical religious event – an ineffable cloud of unknowing.
If the singularity does happen, it will indeed be transformative. But it’s not unthinkable. It’s not incomprehensible. It’s not like we should throw up our hands in helplessness and babble like little babies about how the gods have finally arrived to save us or kill us. It’s not the second coming of Christ.
We can think about it. And, especially if you’re right about its profundity, we ought to think about it.
I’m not aware of any salient constraint coming from computational complexity; what do you have in mind?
More generally, I’m not aware of any good reason to think human-level intelligence marks any physically or computationally relevant threshold, rather than being set by the contingencies of evolution. If so, then there’s no reason to think a process that fairly quickly reaches human-level intelligence won’t comparably-quickly shoot straight past it. Even if AI follows an S curve, there’s no particular reason to expect the transition away from exponential growth to happen to fall around human intelligence and thinking speed.
Yeah, I agree that human-level intelligence isn’t a physically or computationally significant threshold. I expect AIs will surpass it. And I suspect exponential growth could occur a bit beyond the human level. I think there will be ASI.
But I wouldn’t be surprised if it becomes exponentially hard (or harder) to get linear increases beyond human. It was tough (took a long time) for evolution to get to human-level intelligence, though that may be entirely due to irrelevant biological contingencies. But exponential growth is generally very, very hard to sustain in any complex system.
There are already some hints that extremely scaling transformers isn’t producing proportional improvements in intelligence. I think those scaling problems point to computational complexity difficulties for that architecture.
I’m sure new and more powerful architectures will emerge, but I suspect they’ll have their own scaling problems. It might get exponentially or tetrationally harder to find those architectures.
And nobody knows why a bunch of linear algebra generates intelligence, so we just don’t know what sorts of limits may or may not emerge. So I’m generalizing: there are lots of problems in computer science where you can get decent solutions for cheap, but optimizations or extensions get harder and harder. I inductively expect AI to be like that too.
These are deep and interesting issues.
Cars exist, but there are still footraces, and lots of people go running for fun and fitness. Chess computers can beat any human, but lots of humans learn chess, work hard to improve, and enjoy it. Wherever humans exist, some of them will want to develop and use their own personal skills, rather than outsourcing.
So then: if AI turns out to be profitable, rather than all the AI companies going bankrupt, then I predict that while a disturbingly large proportion of the population will be illiterate, and will just let their phones tell them what to do all day, a small percentage of people–maybe 10 or 20%?–will use AI very little because they want to do their own thinking.
People like that will want to go to college and will be excited about writing in blue books, taking oral exams face-to-face, and so on. So there will be work for professors.
Also, there will still be interest in research done by humans, and therefore there will be conferences and other get-togethers for human philosophers who want to do human philosophy, in person, with other humans.
What I’m not sure is what will happen with philosophical publishing. People will still write for their own sake, and they will probably share their writings with others who know them and who trust them not to be providing AI stuff, but a huge increase in AI-generated papers in journals may just eat itself, leading people to stop caring about journals. (No one has chess engines just playing games all day long and posting them to the web, and if they did, few people would spend much time looking at those games. Chess is for humans, and so is philosophy.)
I’m a terrible chess player, butI do think people spend a lot of time looking at chess engines playing themselves. This is where a lot of the newer openings and stratigies have come from, and its been lamented by older generations of chess players because it rewards memorization over creativity.
I suspect that if AI progresses in its abilities and usefulness to the point that a significant number of people offload most of their mental tasks to it, then the number of people who use AI very little because they want to do their own thinking will be comparable to the number of people today who use cars/buses/trains very little because they want to do their own running/biking.
In ten years: I’ll walk into class, open up chatgpt, and say “Please make a 50 minute video of a professor giving a lecture on Gettier and the problem with the TJB definition of Knowledge–scan the room for anyone putting up their hand and answer any question they ask”. Then ill ask chatgpt to write up essay prompt for students to submit essay in response to. The students will use chatgpt to write their essays, and I’ll upload them to get chatgpt to assign a grade. Students will use chatgpt to automaticaly schedule emails complaining that their grade will cause them to lose a scholarship if I don’t raise it. My chatgpt will compose an email explaining why i can’t do this. One friday per month, I’ll send a chatgpt bot to a faculty meeting and 16 or so chatgpt bots will argue about inconsequential decisions the department needs to make.
In 10 years I see academia becoming a jungle and survival of the fittest. AI, MOOC and OER will demistify knowledge as the preserve of the Ivory Tower. The traditional expert accolade from academia will give rise to experts of the masses. Will it be for good or for bad. I think developed economies will not feel the impact that much because many other opportunities exist for the already built knowledge economies to ride on. My fear is that developing economies especially in sub-Saharan Africa, lacking job opportunities for the youth will suffer confidence crises in university education. Confidence in academia will drop, lecturer’s will become frustrated with the monotony as AI makes human creativity and input, so to speak, “redundant”. Job satisfaction will dwindle as supervisors struggle to decipher between originality of works created by students and with the support of AI.
In ten years I believe a higher level of education will be introduced to the general public. Seeing and experiencing the rapid rise of AI and how much progress has been made in the field raises the question of whether or not we will ever be able to stay ahead of it. Seeing this in our time, it does introduce a an interesting point as to whether or not people will continue to strive for education and knowledge given that most of everything universities can offer will be available in the palm of their hands.