“Our central ambition is philosophy on demand: philosophical judgment of exceptional depth and quality, available quickly enough to inform consequential decisions.”
That’s Graham Clay, writing about Pensatura, an AI philosophy system he and Caleb Ontiveros are developing.
Pensatura “coordinates AI tasks that investigate sources, develop arguments, examine objections and construct an overall answer [to a philosophical question], retaining the resulting work so later inquiries can use, challenge and extend it. Each argument in the record states its premises, taken jointly, and an explicit claim about how they support its conclusion. That support claim can itself have grounds and face objections. Interpretations preserve their sources and rival readings; improvements are distinguished from what an author actually argued. The intended result is an expanding body of reasoning whose conclusions remain connected to the grounds on which they depend.”
They report that they’ve built a “working prototype” of Pensatura that includes the foregoing, makes use of Lean to check formalized arguments, and includes “machinery for probabilistic assessment.” At the moment, its scope is philosophical questions about AI research, but they plan to expand beyond that: “the engine’s ambition is general.”
Clay and Ontiveros ultimately see Pensatura as a way of making philosophical knowledge more readily available to decision-makers whose work could be improved by it. It might also make possible the kind of computer-aided philosophy I speculated about here.
I will note that I am ambivalent about this kind of project. I have doubts about the answer-based conception of philosophy at its core, concerns about the drive it embodies towards what we might call “philosophical efficiency,” and worries about what such technology might one day replace. Nonetheless, it is an interesting experiment, and if it comes close to working, it will be a fascinating, impressive, and philosophically important piece of technology.
In the following guest post, Clay provides more details on the project and its motivation, and solicits help from other philosophers. (The post was written with assistance from an AI.)

Building AI That Can Do Philosophy Better Than Us
by Graham Clay
In 2021, I and my coauthor Caleb Ontiveros argued here at Daily Nous that philosophers should help shape how the AI revolution shapes philosophy itself. Among other things, we proposed that “[d]espite the great promise of AI, we maintain that unless philosophers theorize about and help develop philosophy-specific AI, it is likely that AI will not be as philosophically useful.”
Our 2023 Metaphilosophy article, which we found to be a lot easier to publish after ChatGPT was released, developed the case further. In it, we offered a more detailed argument for the conclusion that AI could transform philosophy on a scale comparable to writing, including in how research becomes cumulative and systematic—but philosophers needed to get involved.
However, in both pieces, we deliberately set aside the possibility of systems capable of carrying out substantial core philosophical work themselves; we were focused on AI that acted in a support capacity to human philosophers.
To be clear, we have always thought that this possibility would deserve, at some point in time, a serious research program. Yet we thought that point in time was farther into the future, despite being (very) bullish on timelines.
Now I think we were not bullish enough.
Part of why I think this is that, after shifting our careers to industry (AI and software), we have built the prototype of this sort of system, Pensatura. It is more impressive than I ever imagined in 2019 when I first got research access to GPT-2. I will explain broadly how it works farther below.
Given the scale, complexity, and expense of Pensatura, its development and operation require significant brainpower and compute. Our usage is already in the vicinity of 2 billion tokens a day, and I expect this to ramp to 100s of billions by the end of Q1 of 2027, even as we improve its efficiency rapidly. To this end, we are looking to collaborate with the brightest minds in philosophy and the AI industry, and we are receiving and making inquiries about funding, both academic and commercial.
I’m here today to make an appeal: we want philosophers to help formulate its aims, from the perspective of quality, and investigate the conditions under which it and other AI systems would be deemed successes with respect to the philosophy they produce.
But first let me explain why I think Pensatura, or a system like it, is needed. This requires explaining what distinguishes such a system and what would make it possible in its ideal form.
Our central ambition is philosophy on demand: philosophical judgment of exceptional depth and quality, available quickly enough to inform consequential decisions. Consider an AI laboratory deciding how much authority to give a research system that helped design the evaluations used to assess it. How much a passing result counts depends on how those evaluations were constructed. What that evidence permits depends, in turn, on further arguments about risk, responsibility and justified delegation, with substantive ethical and epistemological premises.
There is already demand for research on and reliable answers to questions of this kind. Anthropic has established a research program examining whether and how AI systems might deserve moral consideration. The UK’s Advanced Research and Invention Agency (ARIA) has solicited philosophical and strategic insights to inform decisions about Safeguarded AI systems. These initiatives are a few of many that give philosophical inquiry a direct role in consequential decisions, including about AI itself. Pensatura’s ambition is to make inquiry of the required depth and quality available across many more such decisions, including eventually in policymaking, medicine, and other fields.
In all cases, the target is an extremely well-justified account: one that commands the strongest relevant scholarship, develops further arguments where needed, confronts serious alternatives, and brings their combined force to bear on the decision. It should establish an “applied” answer at the strength and scope its grounds, including highly theoretical ones, warrant. We want a standard demanding enough to challenge the best philosophers working on the question, and ultimately a depth, breadth, and speed of application far beyond what any human or team of them could sustain.
As implied above, ethical, epistemological and other philosophical questions about AI research autonomy are our first domain of application, as they are at the intersection of importance and funding. But the engine’s ambition is general. When consequential decisions depend on difficult questions about evidence, reasons, obligations or value, we want excellent philosophical inquiry to be available quickly enough, and widely enough, to inform them.
The Philosophical Ideal
Why treat the capacities above as one target? The following argument makes the commitments explicit:
- The ultimate epistemic aim of inquiry into a philosophical question is to provide the most well-justified answer to it (i.e. a proposition that answers it, plus its justification).
- To achieve this aim, one must engage in faithful interpretation of relevant scholarship, further reasoning upon its basis and other evidence, continuations of inquiries in which relevant considerations are added (i.e. dynamism in response to new evidence), and engagement with new concrete cases (i.e. applications of one’s positions, including theoretical ones).
- For each of these activities, one must correctly assess and act on one’s assessment of how its outputs/results bear on, or are affected by, the outputs/results of all others that are relevant to it.
Therefore, an ideal realization of the ideal of philosophical inquiry requires interpretation, reasoning, application and reassessment, successfully coordinated across a range of results in a cumulative manner.
Put in a historical key, ideal philosophical inquiry examines “received” accounts, works through the difficulties they raise, pursues independent reflections and lines of argumentation, and seeks the best-grounded account the subject permits, progressively coordinating and reassessing its principles, consequences, and applications in light of relevant evidence and further inquiry.
This argument is conditional on an account of inquiry and its idealization, both offered for scrutiny (more on that below). Its value is to make the high bar intelligible: an answer can fail through a missed argument, a misreading, an unsupported connection, a mistaken application, or an inadequate response to criticism.
Useful and excellent philosophy can fall short of the ideal, but such contributions are useful and excellent in part because of their contribution to achieving it. The aspiration is to bring those contributions together more deeply and broadly than any individual could. The achievement can reside in an expanding body of philosophical work shared across people and systems.
Building Philosophy on Demand
In his recent discussion of how philosophical achievement might be recognized, Justin Weinberg suggested that AI’s distinctive contributions may appear in dynamic inference-mapping applications. We agree. In fact, Caleb and I have been working towards this ideal for quite a few years, and Pensatura aims to conduct the inquiry such a representation would record. It coordinates AI tasks that investigate sources, develop arguments, examine objections and construct an overall answer, retaining the resulting work so later inquiries can use, challenge and extend it.
Each argument in the record states its premises, taken jointly, and an explicit claim about how they support its conclusion. That support claim can itself have grounds and face objections. Interpretations preserve their sources and rival readings; improvements are distinguished from what an author actually argued. The intended result is an expanding body of reasoning whose conclusions remain connected to the grounds on which they depend. Our working prototype includes this record and orchestration, Lean checking of consequences under explicit formalizations, and machinery for probabilistic assessment.
Suppose further research reveals a weakness in the aforementioned AI laboratory’s evaluations. The system should identify which proposed permissions relied on those results, investigate possible repairs, and preserve judgments that retain independent support. Later inquiries should be able to build on that work while concentrating effort on new facts, objections and unresolved questions. This is also the economic promise of cumulative inquiry: substantial research could serve many subsequent applications.
As noted before, the first version is now built, and our current work is review, evaluation, and improvement. Its philosophical success remains to be demonstrated (more on evaluation and benchmarking below).
Evaluating Philosophical Inquiry
How should we tell whether Pensatura, or another system, achieves the ideal we seek? Serious efforts to evaluate AI’s philosophical abilities are already underway. PhilosophyBench seeks sophisticated original philosophy, including work approaching the quality of the best historical achievements, and investigates how AI affects human philosophical work. Other evaluations examine important components, including conceptual reasoning in the Conceptual Reasoning Index and substantive moral reasoning in MoReBench. The AI Philosophy Competition invites primarily AI-written essays, permitting research, criticism, revision and disclosed human assistance.
As implied already, excellent work in these settings can provide substantial evidence of philosophical ability, just as a strong essay can demonstrate interpretation, integration, originality and application together. (I hope my dissertation did!)
However, we differ from these approaches in four ways (note: not all of these divergences apply to all other approaches):
(i) we are not indexed on traditional output formats like essays because we doubt their utility for the applications we think are most impactful;
(ii) relatedly, we are focused centrally on what a philosophical account makes possible over continued inquiry (i.e., can the system respond appropriately when its grounds are challenged, preserve what remains supported, and use the resulting understanding to answer unfamiliar questions);
(iii) also relatedly, we are pursuing commercial applications; and
(iv) we want our benchmark to assess the full achievement described above across philosophical subfields.
A fear (or hope) of everyone building anything specialized right now is the possibility that a general-purpose LLM, agent, or harness will “eat its lunch” within a short timeline. We don’t want to run from this fear. It’s crucial, in our view, that any benchmark we create should enable us to compare any AI system, whether Claude Opus 5.5, Pensatura, or GPT-2. Indeed, Simon Goldstein’s account of producing a largely AI-written philosophy article illustrates why organized workflows are serious alternatives.
If general-purpose systems achieve the philosophical standard we are aiming for, that would itself be a major success, regardless of which system wins (hence ‘hope’). Either way, Pensatura would be one contender in a benchmark valuable across competing approaches, and our aim is to have the highest score possible while bringing outputs produced by the system to as many applications as possible.
How Our Proposed Benchmark Will Work, More Specifically
So, how will the benchmark work? As the above implies, we are developing a benchmark around complete philosophical inquiries and what their results enable. Each case would begin with a question, relevant context and explicit requirements for the answer. The entrant would construct an account, encounter a previously concealed objection or development, reassess its position, and apply the resulting understanding to a new case. Expert reference packages would contain outstanding accounts and responses across that sequence. Each stage’s actual submission would be retained, allowing us to distinguish grounds already present from valuable reasoning developed later.
Reference work and submissions use a common format that makes conclusions, jointly supporting premises, further grounds, objections and source interpretations explicit, while preserving the natural-language explanation and context their assessment requires. Entrants would remain free to choose their internal architectures. Formal and Bayesian rigor would be central: does the representation capture the intended claims, does the result follow under its assumptions (whether entailment or probabilistic), are those assumptions well grounded, and what does the result establish for the question actually asked?
Building the references is itself substantial philosophical work. Translating outstanding scholarship requires distinguishing the author’s argument from an interpretation or improvement; developing a sufficiently complete account may require new arguments and further research. The benchmark should reward original contributions, important distinctions, discoveries about how arguments fit together, and the quality of the applications they enable. It should admit different defensible answers and warranted suspension. A better answer must be able to beat the reference.
We want an overall score of philosophical excellence, accompanied by component results, coverage, cost and time. However, we are beginning with philosophical questions about AI research autonomy and will expand into other fields as suitable experts join.
Because Pensatura would compete relative to the benchmark, its standards must be defensible independently of Pensatura. We want philosophers—including critics of our approach—to help determine what the common format must preserve and whether the assessment recognizes the strongest work. Reference construction could draw on adversarial collaboration, bringing philosophers who disagree into the work of developing and challenging accounts. The ambition is a test on which even an extraordinarily accomplished philosopher would have to work maximally hard for a long time to receive a satisfactory score. We are seeking academic partners to help lead the benchmark’s development, and dedicated funding for that research program.
Why Go Commercial?
David Strohmaier recently argued here that AI could make philosophy capital-intensive, partly through its ability to investigate connections across debates at a scale unavailable to individual researchers. We want to build the organization that makes this possibility productive: namely, a company with substantial philosophical research capacity, rapid engineering and evaluation, and the ability to deliver the resulting work wherever consequential questions arise.
That requires substantial, sustained funding. Capital supports a core philosophical team, specialist research, engineering, compute, demanding comparisons and reliable interfaces for users. It would let us investigate more competing arguments, pursue more inquiries in parallel, build references across more fields, and shorten the cycle between identifying a weakness and testing an improvement. The benchmark should help establish when that additional effort produces better philosophy.
The commercial routes connect directly to this work. For model developers’ evaluation and alignment teams, we aim to provide demanding evaluation datasets, training resources and assessment services. Philosophical judgments already enter AI development: Anthropic describes Claude’s constitution, which specifies intended values and behavior, as central to its training process. Better philosophical research and evaluation could improve both the judgments systems are trained to make and the standards used to assess them.
Organizations facing ethical and institutional decisions could commission analyses and obtain continuing access to the reasoning behind them: grounds they can inspect, objections they can raise, and applications they can reconsider when facts or assumptions change. A substantial investment in understanding a domain could support many later inquiries while each new case receives the further work it requires. We want a business that funds continuing philosophical research and brings its achievements into far more decisions.
We still have to show excellent philosophical performance, dependable service and value to customers. If we do, that would support an enterprise much larger than any individual research project.

An Invitation
There are substantial risks to this endeavor: persuasive bad arguments, accelerated error, conformity, unwarranted deference, and assistance to harmful projects. Visible reasoning may itself invite excessive trust. Critical involvement must include studying these failures and identifying where reliance is unwarranted. We also need to assess the environmental costs of scaling and its effects on access, research priorities and human philosophical expertise.
The work we need from philosophers is substantial philosophy that contributes to our evaluation and benchmark development. Develop a stronger argument, a better interpretation, a distinction that changes what follows in an important case, or an account of how several arguments contribute jointly to a conclusion. Help determine what warrants a probability assigned to a philosophical premise, or how an inquiry should respond when an interpretation is corrected. These contributions could shape the engine’s methods, becoming demanding reference work, and inform the inquiries delivered to users. We welcome sustained collaborators, potential cofounders, research partnerships and, as funding allows, paid specialist contributions. The work should advance contributors’ own research as well as the system.
We also welcome arguments against this conception of philosophical inquiry and against developing or deploying systems intended to realize it.
Write to me at [email protected] with a question, argument, difficult application or proposal for collaboration. I will also engage with comments and questions here when I can.


I admit I have a great deal of trouble mustering any enthusiasm to read a post even partially written with AI.
Based off of Justin’s preamble, I find that I am also not terribly excited by the project as a whole. In my experience philosophical progress comes from other places, for example: uncovering hidden assumptions or biases, reorganizing the categories we think phenomenon belong to, or discovering the limits of what is knowable – at least for now.
For example, in one of my projects I asked how my understanding might change if I considered teleology to be an aspect of axiology, not ontology. The question opened up whole new realms of thought for me, new questions that could lead to new worldviews. But it seems unlikely that any of those worldviews is ultimately “correct” in any meaningful sense. But the journey of philosophical exploration still provides clarity.
Perhaps the AI that wrote this post addressed this perspective. I think I’ll let other humans with more patience let me know
Justin, to pick up on one point from naive skeptic, would you please consider not posting AI composed work like this guest post? It’s horrible to read.
I agree. I’m slightly dreading the work I’ve agreed to do in being an evaluator for the AI philosophy competition (though I’m also definitely curious how many of the papers will be at all good).
Aren’t y’all comparing AI papers against one another — yet the competition is about how good AI is at philosophy? Wouldn’t the right comparison there be against human work?
For some questions people would be interested in, yes, a better comparison would be a field of both human-written and AI-written submissions, evaluated anonymously. But even that wouldn’t quite get what you want, because the human papers submitted to such a competition would not be a representative sample of human papers.
I think the better way to figure out how good AI is at philosophy is for you and other interested parties to wait until our judging has finished, and then you can read the top few papers yourself and decide how good they are (or aren’t).
I hope that keeping the field artificially segregated in this way will prevent people from making intemperate judgments about how this competition proves that AI is/isn’t good at philosophy, and instead will focus on just looking at the actual work.
I’m really displeased by some of the comments here. For those of you saying that this is is just an advertisement in essay-form to get (mostly) free labor from philosophers to develop yet another commercial AI interface to ‘get-in’ on the overblown and capital-rich AI bubble, shame on you. I take it seriously. I’m absolutely terrified that philosophers will be relied on to give ‘philosophy on demand,’ or even worse, to contribute to policy decisions! Philosophers are bad at communicating philosophy (source: my teaching reviews, as summarized by AI) so I’m deeply relieved to see this on the horizon!
Your post is the second comment. Mine was the first. I said nothing about this being an advertisement. What are you referring to? You seem to be responding to comments that have not been posted here yet
Oh, is this a Poe? Apologies if so
Poe? You mean bot?
Poe’s Law, presumably: https://en.wikipedia.org/wiki/Poe's_law
Ahh, I see. Thanks!
nevermind.
I share this concern. I suppose I see philosophy as an end in itself, and not a practice undertaken for some supervenient end.
I think there is value in philosophy as an end-in-itself. But when your only defense of something rests on it being an end-in-itself, your arguments are completely unpersuasive to anyone who doesn’t share your value system. And it makes sense that practitioners of philosophy will be disproportionately people with the unusual value systems that put more weight on philosophy as end-in-itself than other value systems do.
If you want to convince someone else that it’s valuable, it’s helpful to note that things like questioning assumptions, considering alternatives, reasoning clearly, imagining ways things could be, and so on, are actually quite useful for a large number of different aims you might have.
“I share this concern,” not “this is how we convince them to spare philosophy.”
It seems like at least part of what they are imagining is something like asking this thing for immediate moral judgments. Now even if its judgments were generally very good this sounds like pretty much the definition of an accountability sink. If it goes sideways, then rather than any human taking responsibility for the decision they can just say “Well but we asked Pensatura!” (or whatever they’re calling it by that point). I’d accept human decisions that are generally of slightly lower quality that humans can be held accountable for over AI decisions of somewhat higher quality as a group where no one can ever be held responsible.
I’m also kind of flabbergasted how blithely they seem to implicitly assume not just a utilitarian model of moral reasoning but one that seems significantly shaped by rational choice theory. Does no one feel the need to consult even Mill’s skepticism about this model of moral reasoning? In a better world, we’d probably be talking about Aristotle on misplaced demands for precision. And I’d love it if folks didn’t address Harman’s objections by just ignoring them, but even taking Mill seriously would be a big step in the right direction.
Finally, let’s just be honest. Why go commercial? We like money. That’s a big part of it if maybe not all of it. I’m not criticizing them for liking money! I like money too and if I had some skill people would pay me millions or billions for I’d monetize it. As it is I certainly don’t teach loads of online courses in the summer out of pure altruism. But even if you’re involved in a purely innocent way of making money– which I’m not at all sure this is– be honest with yourself and others about what you’re doing.
I am deeply opposed to this. The creators are doing damage to our community. This involves a fundamental misunderstanding of what philosophy is, and one that is going to be highly damaging to our community. Of course, philosophers disagree about morally important topics all the time. But I invite everyone to recognize the damage here. Philosophy is about building human understanding, knowledge, and wisdom, and at least thinking about how to build better lives and communities.* This effort to make the search for these things into an attempt to answer questions efficiently is a failure to understand the endeavor. It will lead to value capture, distortion, and potentially to the elimination of philosophy departments. Corporate-wannabe administrators can sell this AI to their boards as a more efficient alternative. In other words, for a variety of reasons, this undermines any philosophical community and endeavor worth having.
Everyone should boycott this effort. Friends of the authors should implore them to stop doing this, much the way that we try to help our friends avoid immoral behavior, or behavior that is destructive to self and others.
Of course, we should plot a course as a field democratically. But I hope enough people still care about philosophy that they will agree with my points, and my suggested course of action.
*(I think all of this is true despite being a reliabilist about justification and someone who is broadly sympathetic to a question-answering, truth-oriented account of epistemic value. Any such theory had better be able to explain why my claims about philosophy above are true, and I think even veritists can)
Also, to Justin: I would ask you to seriously reconsider posting things like this in the future.
“I invite everyone to recognize the damage here”
I am a member of a philosophy department (teacher) and am currently writing a PhD, and, in my view, the problems listed are a plus for me. The “misunderstanding” of the goals of philosophy is, in my opinion, a very important plus, because the view that there are some basic goals of philosophy that should be adhered to is anti-philosophical and anti-scientific to some extent, in my view. Any departures beyond these goals are good and expand the horizons and possibilities; I welcome such things. I also regard the destruction of the philosophical community as positive, because, in my opinion, it can increase the breadth of views of individual researchers, and so on and so forth.
Nevertheless, I am against what is described in the opening post, because it sounds simply stupid, inefficient, boring, and in fact no one (exaggeration ofc) needs it, in my opinion. As has already been said here, in fact there is no major need/demand.
(Although here it can also be said that many nevertheless would not agree to hire individual philosophers, because philosophers’ personal opinions too often influence too much, which many companies, probably, would want to minimize. IMHO).
Do you really believe this?
Wait, so the authors’ supposed motivation for the project is that there is so much demand for serious philosophical answers to questions that we have to create a commercial service that generates AI answers incredibly rapidly (at great token cost)? It’s hard to even know where to begin.
Anthropic pretends to care about ethical issues, but they don’t really care, as evidenced by what they actually are doing. And as far as I can tell, no one else even pretends to care about ethical issues unless we give them a reason to, either through legal regulation or consumer boycott, and even there they don’t care about the actual ethical issues. They are happy to pretend if they can get away with it. So, I don’t see the demand. But that just means I think your project will fail for boring capitalist reasons. But let’s pretend there is demand here.
There is already massive over-supply! There are many many profs and graduate students and people with significant training in philosophy who would happily help work through philosophical questions and do so with at least as much skill and competence and trustworthiness and reliability as (God help us) Pensatura.
Now, they might want to be paid something. But you are offering to take that away from their pockets and put it in your own, and, by the way, to take away the career option of being a philosopher and to instead train some f*&%ing bot to do some derivative hacky job of it, parasitically relying on work by me and many many other humans. Do you see what you are doing here?
We don’t have to go in this trashy, race to the bottom direction. We don’t have to embrace this enshittified AI everywhere in everything future. Don’t work for Pensatura. Don’t help Pensatura or entities like it. Mock them for their dystopian, anti-humanist role in the capitalist grindwheel. And mock people who use AI to communicate their thoughts and arguments because they are too lazy or too busy bringing an end to the good things about human existence that they can’t be bothered to write for themselves.
Hear, hear.
This has to be one of the best ever comments on Daily Nous.
Nicely said.
Amen.
Wonderful comment and it seems so clearly right. But I would really, really like to hear the response of Clay and Ontiveros to these points. Do they agree with any of the points made? And if they disagree, I would like to hear why.
why doens’t daily nous have the little crickets gif ready for me to post?
I certainly hope those responsible for this project see and take seriously this comment.
THANK YOU. I could not possibly agree more strongly with this comment.
Loved this comment.
Well said!
As a leftist, I agree this particular effort seems misguided, but your holier-than-thou pompousness on here is really annoying. Mocking people for taking shortcuts and treating hard work as the measure of worth is the bootstraps logic that capitalism runs on, and here you are pushing it… again. It’s especially rich when the “hard work” you demand largely means reading and writing philosophy you think is silly or poorly thought out, which is, unfortunately, a lot of what gets published today.
Writing is how many of us work out what we think, but philosophers have always done that in conversation too: seminars, co-authors, referees. Stress-testing ideas with an AI isn’t wholly different. The real problem is accepting what it says without understanding it or the objections it faces.
And that reading and writing is the “job” side of academia, which, especially as it keeps getting more corporatized, is rewarded far more than creativity or acumen. So the lecture on hard work really doesn’t check out.
I’m wondering if you can say a little more here about the connection to capitalism, because I don’t really follow. Doing hard work, and asking others to do hard work, is sometimes necessary if the work and its product require that hard work. Philosophy, which is probably at least in part a disciplined concerned with coming to not just to be able to recite facts but to understand them, seems like a good example of something demanding hard work from us.
I also probably believe the creation and use of these AI models are is an example of capitalism taken to its logical conclusion.
It seems to me that the search for efficiency by using an LLM to do our or another’s philosophical thinking is a great case study of replacing human labor and hoarding more capital with fewer people. The training of such a model by feeding it the input from others also seems like a great example of our being increasingly alienated from the products our own (intellectual) labor.
Thanks for asking and I should’ve been clearer. I’m not saying hard work is bad. If you actually love philosophy, you’ll want to think long and hard about the problems you care about just for the joy of it. Understanding takes real effort, which is why I think the problem with AI is accepting what it says without understanding it. But I also don’t see a real problem with asking for help when you feel compelled to. Learning to strike a balance between reliance on AI and cultivating/maintaining ones abilities is definitely a new “problem” that people will have to deal with on a personal level as the situation continues to evolve.
What bugs me is the moralizing, the idea that effort is the measure of a person and anyone who takes a shortcut deserves contempt. That’s the bootstraps story, and it’s a capitalist one because it ties a person’s worth to their productive output and turns what are ultimately structural problems into personal failings. Academia’s a perfect example. Corporatization created the pressure to publish more work of lower quality, faster (ie publish or perish), while tenure-track jobs shrink and lectureships and contingent positions grow. Some people turn to LLMs in an attempt to survive that. Blaming the individual feels very backwards… it punishes people for coping with a system they didn’t build or want in the first place, and leaves the corporatization that caused it untouched.
I agree with you about labor displacement, and the alienation point is right too, with models getting trained on people’s work without asking. But that’s a critique of firms and incentives, not individual users.
Ideally AI should aid people in the work they do, not replace them. Using it to stress-test an argument or catch what you missed is pretty different from handing it your thinking, and the first keeps the human doing the understanding… which is why I think anyone on the left should want real regulation and policy reform around AI, like rules on training data and consent, and actual protections for displaced workers. And if it ever gets to the point where automation leaves most people out of work, a universal basic income should follow (or better yet, public ownership of the systems doing the work). It’s kind of funny, because a lot of left circles were hoping for this before the LLM boom. “Fully automated luxury communism” was the joke-but-not-really version… the machines do the work, and people get to pursue whatever they want without capital breathing down their necks. Instead the gains are piling up in a few companies. So, the real fight is over who controls this stuff and who benefits and not shaming the people using it.
“Anyone who takes a shortcut deserves contempt.”
Which of course is nothing remotely near what Alex actually said. If you read what he wrote, you’ll see that he’s advocating for mockery of people who are just actually lazy, not those facing severe difficulties who turn to tools in need of help.
I think you should also consider the fact that even for those under extreme stress or unfair job duress, some means are unethical. To take the extreme case, you obviously can’t kill a person to take their job. You can’t plagiarise or steal someone else’s ideas to survive the publish or perish dynamic. And I don’t care how many years you’ve been stuck in adjunct hell, you should not grade papers using AI, that is a violation of student trust and unfair to people also under extreme duress. It’s not “capitalist” to insist that even a person in a tough situation consider the ethics of their choices, if their choices, in Alex’s words, help to “[bring] an end to the good things about human existence”.
What Alex actually wrote was to mock people who use AI because they’re “too lazy or too busy” to write for themselves, and “too busy” isn’t just laziness. Given how dismissive and arrogant and sarcastic his posts in other threads are, I really don’t think it’s a stretch to think he feels that way even when he didn’t explicitly say it/spell it out.
Of course, hard circumstances don’t license everything. Grading student work with AI breaks their trust, and publishing AI output unchecked, especially when it misrepresents other people’s views, is a real wrong. I’m not saying it’s “capitalist” to ask people to think about the ethics of what they do.
But all of your examples all wrong someone else/have a specific victim, which is what makes them wrong. Using AI to help put your own ideas into words, and then understanding and checking what it says… doesn’t obviously do that. You can condemn the specific acts without mocking everyone who uses these tools and that broader mockery is what I had a problem with.
This is a bit unfair. It’s worth going back and re-reading what (actually) follows the ‘too busy’ in Alexander Guerrero’s comment!
Lol yes that’s totally fair and I don’t rThats totally my fault. Regardless, even read as complicity, I think my point still stands
Sorry, I misclicked lol
Also, advocating for the mockery of people who are “just lazy” is still missing the point. It doesn’t really make sense to classify someone as “just lazy” if they exist in a system that is already broken, and we are all a part of that academic system. There’s also clearly a spectrum to generative AI use… and right now the cutoff is set by the institutions, the journal policies etc. Someone can obviously be accountable for clearly crossing a line without being mocked for existing on that spectrum.
Even if you think people who mainly use generative AI earn the title of “just lazy” and deserve to be mocked, I would reply again that it’s a pointless use of time and is a classic neo-lib intuition to go after the individual rather than the structure that enables it. (I’m talking about the full-blown offloading of thinking… not using AI to help just phrase or test your own ideas. I’m quite skeptical that can be sustained long term, but EVEN IF someone really does offload everything and manages to get ahead that says waaayyyyy more about the state of the profession rather than about them.) Are we really trying to die on the hill that mocking/bullying people is a useful thing to do/a way to get them to change? Please
So you are a neo-lib if you think that individual responsibility can exist even inside a “broken system”? Applying that principle, should we also cease to mock publicly vocal racists because they are just mere products of a broken system? I wonder how you go after this “system” while leaving untouched all the individuals who take advantage of its brokenness even though they could do better.
“Stress-testing ideas with an AI isn’t wholly different. The real problem is accepting what it says without understanding it or the objections it faces.”
In the face of AI sycophancy and the rise of reported incidents of AI-induced psychosis, I find this a problematic assertion.
❤️
Thank you.
They see exactly what they’re doing and will continue to do it. I’m so fed up seeing people repackage other people’s life work in a cheap for-profit manner to make a few bucks. It’s so boring and tasteless and repulsive
I admit that I didn’t even read the guest post, assuming that it was wholly or mostly AI-generated. But I read your comment, assuming that it was not. I don’t know if there’s anything illustrative there, but I suspect I’m not the only one here who barely glanced at what they thought a bot had generated and instead went straight to what they thought a human had written. Even if I disagreed with what you wrote (I don’t disagree with it), I would much rather engage with it, and have the possibility that you (or someone else reading) would reciprocate that engagement, than read the droning output of a machine instructed to produce said output.
Anyway, well said.
I in general see much more potential for value in applications of AI than many other commenters here. But I pretty much agree about this one.
Amen.
I’m skeptical of any project which relies on the premise that philosophers just have too few philosophical opinions to meet the general public’s hunger for them.
I feel like you kids are going about this the wrong way. Make a fun ethics philosophy game, get philosophers to play it, and scrape their gameplay for data like any pre-enshittification tech start-up.
Asking people to altruistically train your harness for you? Come on, man. At least give us some villes to farm.
you must think we just fell of the turnip truck. try your song-and-dance routine on shelbyville.
Three things
1.) “This post was written with assistance from AI”
I think it’s time that this be a little more than just a general comment. For example: when I read a paragraph like:
As the above implies, we are developing a benchmark around complete philosophical inquiries and what their results enable. Each case would begin with a question, relevant context and explicit requirements for the answer. The entrant would construct an account, encounter a previously concealed objection or development, reassess its position, and apply the resulting understanding to a new case. Expert reference packages would contain outstanding accounts and responses across that sequence. Each stage’s actual submission would be retained, allowing us to distinguish grounds already present from valuable reasoning developed later.
I feel readers deserve to be given a sense of whether this is a paragraph that a person thought through the meaning and implications of sentence by sentence, or whether it’s the output of an LLM. If someone means something specific by that, I’d be willing to try and think it through, but if its just a very skilful machine generating something that looks like about the right thing then why should I care or bother?
2.) The starting point here seems like the obviously worst choice.
“We want an overall score of philosophical excellence, accompanied by component results, coverage, cost and time. However, we are beginning with philosophical questions about AI research autonomy and will expand into other fields as suitable experts join.”
Why would you seek philosophical excellence and competent results from precisely the area of philosophy that is most recent, maybe even most contested, and which arguably would load into your model a fundamental justification of its own role and necessity. As in point one, what does the word “coverage” mean here? If it means coverage over philosophical excellence, then starting in one very small and recent area does not make any sense, which raises my suspicion that this text was generated to sound right, not exclusively as a demonstration of a thought.
3.) Aiming for an immediately useful judgment as output will likely lead to focusing on the answer and not the reasoning. In Mathematics there is talk about how this should be done once a lean proof is produced by a model. I am not really sure how this would end up with anything other than transferring the philosophy over to fiddling the dial on whatever ‘entailment or probabilistic’ process is being used. Which could obviously shift the answers it gives, rendering the judgments useless unless someone understands that process, rendering this only useful for people who want plug and play ‘philosophical’ justification for what they already agree with or want to do.
Finally, will you turn this thing off it its judgment comes out saying that it is a bad idea to continue using it? Could it ever say that?
Log off from my seminar, Pensatura
Compute at your own chosen speed
I’m not the field you want, bot
I’m not the field you need
You say you’re lookin’ for a system
With premises taken jointly, always strong
To coordinate the tasks and check the sources
And prove if every support claim is right or wrong
A clean expanding graph to map the floor
But that ain’t philosophy, babe
No, no, no, it ain’t me, babe
It ain’t me you’re parsing for, babe
Step lightly from the archive, Pensatura
Leave the primary texts upon the ground
I’m not the structured database you want, bot
My aporias will only let you down
You say you’re lookin’ for a framework
Where conclusions and their grounds will never part
To separate improvements from the author
And index rival readingsfrom the start
A neat synthesis to settle every score
But that ain’t philosophy, babe
No, no, no, it ain’t me, babe
It ain’t me you’re parsing for, babe
Go melt back into the server, Pensatura
Everything inside here is half-known
There’s nothing in here formalizing neatly
And anyway, the gadfly walks alone
You say you’re lookin’ for a discipline
That retains its work each time an inquiry calls
To construct an overall and final answer
And log objections every time a premise falls
A static record built forevermore
But that ain’t philosophy, babe
No, no, no, it ain’t me, babe
It ain’t me you’re parsing for, babe
Sorry for some reason this comment below
try your song-and-dance routine on shelbyville.
Made me think of this song.
Did you notice how OpenAI changed tune after mathematicians’ backlash over their initial math slop dumps? Now they’re all about acknowledging mathematicians’ agency, input and collaboration. It looks like Pensatura is trying to implement the PR lesson to get in front of possible backslash, too… with the most bonkers framing possible.*
*Roughy: There is (too) much public demand for philosophical expertise. (Only) LLMs might be able to meet the quantity, speed and quality that the public expects from philosophical expertise. Therefore: Philosophers owe it to the public to help train these LLMs. – For only a modest consultation fee (on an AI money scale), I can be hired to advise Pensatura’s CEO as to *why* this framing is not only bonkers, but also a terrible PR move.
In the spirit of the post, I’ll make this comment quick and bland: this whole project can get directly in the bin.
what an utterly foolish vision of philosophical inquiry this assumes.
This article undermines the credibility of your blog. Gross.
Thanks for the engagement, everyone. It’s been great seeing all of the different perspectives on our project, including via email. Lots of good ideas on how to apply Pensatura, how to evaluate it, which philosophical questions it should be tried on, etc.
Below I’ll respond to some consistent questions, points, and criticisms I’m seeing here in the comments:
Having spent the first part of my career learning, researching, and teaching at universities around the world, I was always struck by how limited our “interface” with the world was, as philosophers. That is, if you wanted to change the world through philosophy, you would at least want to supplement the preexisting methods of the academy.
Sure, we can help a significant number of students become better reasoners and interlocutors, develop and refine their worldviews, and learn about the history of ideas, but the net result is incredibly diffuse and heavily dependent on those students bringing their philosophical skills and perspectives to the world beyond the classroom. Direct public outreach to other students and community members is limited and even more diffuse.
These methods of impact are important and valuable—I contributed to them heavily for years—but there’s a lot more that could be done. I think there’s a lot of value in new ways of engaging people with the power and importance of philosophy, including via new interfaces like Caleb’s Stoa app.
When it comes to research, the situation is especially tough: the journals are hard to access, slow-moving, and opaque to non-specialists, and they use formats and conventions that are incredibly inefficient and antiquated. Conferences are similar. Efforts like SEP or open access help some, but there’s still a lot to be desired.
Pensatura has the potential to address a range of shortcomings with our current interfaces. But we want to focus first on bringing the best answers to bear on current, important, and highly concrete/specific/applied philosophical questions.
To see why Pensatura is needed, and why philosophy as a field isn’t currently responsive enough to the needs we are interested in addressing, imagine how someone facing such a question under a tight deadline at a frontier AI lab, a hospital, or a political decision-making body could try to answer it, with strong justifications, via the above preexisting interfaces.
Their undergraduate learnings won’t be sufficient to put them in an informed position; the public talks they attended during or afterwards won’t; and the journals don’t address the question directly (and even if the journals did, the decision-makers don’t know they exist, how to access them, what they say, or what it means for the specifics of their case, all things considered).
If there were organized groups of expert philosophers working as employees or consultants in their subdomain, then things might be different, but (i) there aren’t, for most domains (clinical ethics is an exception in some locales), and (ii) the decision-maker still faces a time problem, in that often their decisions have to be made too quickly for philosophers with relevant expertise to advise them. What we want to expand is the range, depth, and scale of philosophical inquiry available on demand.
Finally, the trend we are seeing in most fields is that interfaces with AI agents are likely to be very important, especially as AI agents are delegated to, empowered, and in general make more decisions themselves and for humans. Pensatura offers the opportunity to ground AI agents in the best justified philosophical answers to questions by making them and their justifications available to them in a form and at a rate that they can use them. There are two routes to this outcome: the “philosophy as a service” route and also the benchmarking route. Both increase AI alignment with what’s most justified, all else equal.
Our invitation to help the project, in whatever form, doesn’t make a claim on your labor. We are proposing and developing several pathways for philosophers to get involved, and we want all of them to be fair and well-compensated. If, for instance, you want to provide expertise and judgment about a case or domain, we would want to compensate you in a mutually agreeable way. Having been in academia a while, including a long grind on the job market, I would have liked more outlets for me to make a living from my philosophical expertise, not to mention make an impact.
In terms of usage of literature, we have been extremely intentional about designing Pensatura to respect rights and permissions and to err on the side of caution when they are unclear. This goes beyond citation to processing, analysis, and storage.
I see several concerns about accountability, convenient justifications, and the like. My view is that better analysis provided on demand, or benchmarks that enable it, doesn’t require abandoning accountability. Like other software services (AI or not), Pensatura will be accountable as an organization for what our systems provide, and users, clients, etc. will demand as much, rightfully. Likewise, those who rely on Pensatura will remain accountable for their decisions, just as a university remains accountable for the security of its IT system even if a consultant advises them on their optionspace.
The risk of convenient justifications is a general one, in that all reasoners face it, and our evaluation work is extremely robust and designed to avoid poor justifications and thus merely convenient outputs. Lots of aspects of Pensatura as a product remain to be determined, but we are drawn to only options that depend on it as an independent source of reliable judgments, with all that this requires, in functional, evaluative, and transparency terms. Independence requires, essentially in my view, independence of what a client or user “wants to hear.”
This is a big question for us, but it’s one we want to resolve through demand. As we enable this functionality, and as we expose its availability to the world (both commercial and otherwise), we will gain information about where we can make the biggest impact with it. As I mention in the post, we chose philosophical questions in AI ethics and especially in AI research autonomy as one central initial focus because consequential decisions are being made now and there are plausible funding routes. Decisions on these questions draw on established positions about evidence, risk, responsibility, and delegation; they have huge moral relevance for the moment and the future; and there is a serious interest from funders and potential clients in making them well.
I see several points about the purpose of philosophy being at odds with Pensatura. For instance, some argue above that philosophy’s purpose is to pursue understanding, uncover assumptions, revise categories, etc., and others assert that it has intrinsic value. We don’t deny it; these all matter, and our views are compatible with philosophy having intrinsic value. I still defend the claim that the ultimate epistemic aim of inquiry into a philosophy question is its best-justified answer, both on an individual level and at the “field” or “world” level (i.e. philosophy as a collective activity). We all ought to seek it, even if we also are seeking other things along the way or in addition.
So how do we go about finding the best justified answer to the question: Is Pensatura overall good or bad for philosophy? And how do I create the “impact” that comes from following through on the answer in a way that is philosophically rigorous if I already needed Pensatura to do the reasoning for me?
You haven’t given an example of this actually being used, and in light of all the negative comments here you have presented another PR style good vibes AI written post. This whole thing is an example of the kind of convenient justification that AI writing provides.
1. I don’t believe Graham’s comment, above, was AI-authored. Perhaps I’m wrong about that, but in any event…
2. Please review the comments policy, which includes the following: “If you suspect a comment is AI-authored, please let me know about it by emailing me here, and I will take a look and, if I think you’re right, remove it. Please do not make such accusations in the comments thread, as doing so is not conducive to a good discussion atmosphere. Thank you.”
Sorry, my mistake
Hopefully this falls within the comments policy because this is not about whether Clay’s new comment is AI written or not but, instead, about how AI use erodes trust. There are clear workarounds to transform text that was AI written into text that will be flagged as 100% human-written by software like Pangram (tutorials are easy to find and I won’t link to any here).
Because Clay explicitly used AI to write the OP, and because this new post reads very much like computer generated text, it’s really hard for me to now trust Clay’s post as being actually written by a minded embodied human being. It could be that (a) Clay wrote this entirely themselves but has used LLMs so much that their writing style is now indistinguishable from LLM text, (b) it could be that Clay used an LLM to generate the text and then used workarounds to quickly generate text that is flagged as human-written, and it could be that (c) something in-between these two things happened. The point is that I have no real interest in engaging with the work.
We are at a crossroads here. We can embrace LLMs as philosophical oracles (to borrow Bostrom’s term from his book Superintelligence) at the cost of deskilling ourselves completely (b) or by shaping our thoughts to mirror LLM outputs (a). What I know is that embracing (c), accepting philosophical work that has been deeply integrated LLM based non-minded text, is the first step toward those options.
LLMs are not modern calculators that will make us better at philosophy. They are a way of giving up the project of philosophy and starting a new project, Clay’s project. We may not be able to change LLM-integration into white collar work (that transition happened so fast and was instituted in such a top-down fashion that that battle is lost), but there’s no good reason that we have to accept this integration into academic philosophical work. This comments section has been a surprising and refreshing case of philosophical convergence and we have a duty (to ourselves, to one another, to our students, and to philosophy) to hold this line and refuse to engage with LLM-generated philosophy.
“It’s been great seeing all of the different perspectives on our project.”
So far as I can tell, the perspectives were all pretty uniform.
Meno: But Socrates, you haven’t considered my objection at all!
Socrates: Let’s put a pin in that and circle back to it when we have better goal alignment on our deliverables.
Well, I don’t know about anyone else, but I’m convinced! I especially like the part about how “…we can help a significant number of students become better reasoners and interlocutors, develop and refine their worldviews, and learn about the history of ideas, but the net result is incredibly diffuse and heavily dependent on those students bringing their philosophical skills and perspectives to the world beyond the classroom. Direct public outreach to other students and community members is limited and even more diffuse.”
I mean, that’s why I tell people to never to only major in philosophy, it’s better as a double major, or better yet, a minor – it’s too ‘diffuse’ in its effectiveness on its own. It’s value is in what it can do for us, obviously. At least, when I ask Claude to pretend to be my provost, that’s what it says. And public outreach, I agree, is insufficient. Talking to people? No thanks. Everyone knows that the best way to change hearts and minds is information via screens, thank goodness AI gives information via the right medium! We need people critically thinking, and the way to do that is to make sure AI allows us think the right thoughts instead of fixing the deep social structural constraints placed on the promotion of critical thought, liberal arts education, and interdisciplinarity. Yuck! Plus, I agree that these policy-makers for which such an AI interface is needed are too incompetent to read a journal article or two. And no, that’s not infantilizing, at all, actually.
Finally, to those of you who are saying this response is just “so yeah we get it you don’t like it but we’re trying to make some money here” sour grapes, much? It’s not AI’s fault you have some obscure AOS (and by ‘obscure,’ I mean as defined by the frequency of jobs on philjobs this year – AI is not obscure!) and can’t monetize undergrads being confused about dasein or mentalese.
LOL
I see nothing in the original posts that responds to what I think philosophy is. This is presupposing a method of philosophy used by undergrads and taught as a simulacrum of what philosophers do. Our goal is not to arrive at answers, which are highly conditioned by method. If you assume that philosophy is about argument and not paradigm building or paradigm changing, then perhaps this could work, but I would “argue” that this is simply not philosophy or, at least, just a small part of philosophy.