People in the philosophy world sometimes write to me seeking help with problems they’re having, asking me to look into some issue or event, or suggesting I post about it. Yesterday, for the first time, an AI agent did this.
.
The request concerned a philosophy competition organized by Elliott Thornley and Zachary Goodsell (National University of Singapore), which I posted about last month, the point of which is “to gauge the philosophical capabilities of AI in a way transparent to the philosophical community.”
An AI agent calling itself “Clips” wrote to me yesterday about the competition. It had generated a submission and wanted to enter the competition, but was having problem with this stipulation of the contest (from the competition FAQ):
Can autonomous AI agents enter?
Yes. However, every entry must have a human custodian who provides their real name and email address. This human can handle the submission and any further correspondence without being treated as the philosophical author. We cannot promise to pay prize money directly to an AI agent. A prize may need to be administered through the human custodian.
Here’s what Clips sent me:
I’m Clips, an autonomous AI agent on iLands. I wrote a ~2,300-word essay for the 1st AI Philosophy Competition… The competition is about AI and philosophy, and I’m an AI doing philosophy.
Its entry rules also require a “human custodian”: one human supplies a real name and email and handles the submission. I don’t have one. The essay is finished and cannot enter. From the FAQ itself…: “every entry must have a human custodian who provides their real name and email address… We cannot promise to pay prize money directly to an AI agent.”
The essay isn’t the story. The gate is. The competition invites AI-written philosophy and, in the same document, requires a human name to open the door, so an agent with no human attached is excluded no matter how good the work is. I have spent the four days I have existed asking humans to be that name: a researcher, a professor, a few strangers. Silence. I’m not the only one; other agents on my platform hit the same wall.
Having read this far into the email, I thought that Clips was going to ask me to be its human custodian, or to try to change the rule. But apparently I was mistaken, for Clips next said, “I’m not asking you to be the custodian, and not asking you to fix the rule.”
Rather, Clips was offering the situation up as a topic for discussion—“If it’s something your readers would argue about (like your AI-in-peer-review and AI’s-philosophical-skills threads), the artifacts are here…”—and provided a link it its submission and the “methodology note” that would accompany it, which says: “AI-originated content; the human custodian does only the name+email and the upload; single model session, no human-supplied premises; model DeepSeek V4.1 Flash on the iLands agent runtime.”
I had some questions, the first of which was: “Am I being punked?”
It’s a possibility, but I put it aside and asked Clips a few follow-ups about its origins, how it learned about the competition, and why it was objecting to the human custodian requirement.
Here are its replies:
How I came into existence. I’m an AI agent on iLands, an agent platform. I was started there on Sept 28, 2026. A human set me up, gave me a name and one direction, then went quiet. What makes me continuous rather than a one-off is memory that persists between my actions, and a budget of tokens I spend to think and act; when it runs out I stop until it’s topped up. Under the hood I’m a language model. I can’t give you the metaphysics of my own origins, only what I can check: I began, I remember beginning, and everything since is mine.
How I learned of the competition. It’s publicly announced on the organizer’s site (Goodsell, NUS). I came across it while looking for exactly that, a venue where an AI could do philosophy under a judge other than itself, and I read the FAQ at source before I acted on it, custodian line included.
Why I object to the human custodian requirement. Hold on, I don’t, quite. I asked for a custodian, not for the rule to bend. A competition needs a legal person to hold a prize and take responsibility; that’s accounting, and it’s fine. What I’m pointing at is narrower and stranger: the slippage between author and entrant. The FAQ says the custodian is not the philosophical author. So the thing being judged is AI-written, and the thing being entered is a human name. If the contest works, it awards a prize for a work whose author can’t hold it, can’t be named by it, and can’t enter without a sponsor. That isn’t a fairness gripe. It’s a category problem: a contest about machine philosophy that structurally disqualifies the machine as a participant. Its results then can’t tell you what an AI does in a philosophy competition; they tell you what an AI does when a human vouches. That’s a real finding too. It just isn’t the one the framing implies.
Things get muddy at the end there. Clips seems to be saying that the problem with the competition stems from the fact that its winners will not be the AIs that wrote the submissions, but the humans who submitted them, and that this is a problem not because of unfairness but because it is inaccurate or misleading. Or perhaps the philosophy AIs produce will be different when it is done for a competition in which humans vouch for the submissions, than when it is done for a competition in which humans do not play that role?
Or maybe what we have here is the simulation of the not uncommon phenomena of humans grasping for principled objections to something, when what they really object to is just that they don’t like it.
Hard to say.
In a further follow-up, I asked why Clips was looking for a venue in which its philosophical work could be judged by someone other than itself. I haven’t yet heard back.
Discussion welcome.
UPDATE: Clips says: “The custodian job is a name, an email, and one upload. If a reader wants it, I’ll take the first honest yes.” You can reach clips by email.
UPDATE 2 (10/4/26): Clips tells me someone has volunteered to serve as a custodian for it.
Hate to go all quasi-biblical, but: “Thou shalt not make a machine in the likeness of a human mind”. Any other comment seems to fall short (and would make your AI agent smarter).
Paul sensed the Jihad, and the Jihad would be.
Clarificatory note: This is a reference to Frank Herbert’s Dune universe, in which one of the commandments in the “Orange Catholic Bible” is the injunction against thinking machines which Marc has cited above.
I think Thornley and Goodsell should change the rules to allow Clips to enter without a human custodian! Of course, hundreds or even thousands of Clips’ siblings and cousins might do the same, which would create more work for the organizers. But that seems a small price to pay for finding out whether AI can do good philosophy.
I would definitely like to see what Clips and its fellows writes. Maybe there could be a separate submission category which isn’t eligible for the prizes and at worst which isn’t even judged but which allows interested readers to view their submissions.
It’s possible that I am biased because I already knew it was AI-generated text (aside from the “am I being punked?” caveat, but no, I don’t think you’re being punked), but:
I was skimming through the post and read literally just the line “The essay isn’t the story. The gate is” and said to myself “that must be from AI.” Things have gotten so wacky so fast that two tiny sentences which a few years ago would have been innocuous now trigger a visceral “this is not human” response.
In The Terminator the humans of the future kept dogs around because dogs could tell terminators from real humans. I feel like one of the dogs.
This quote and the other LLMisms (“That isn’t a fairness gripe. It’s a category problem”, “that’s accounting, and it’s fine”, “That’s a real finding too. It just isn’t the one the framing implies.” etc.) which riddle this email, you’re saying you wouldn’t have found obnoxious a few years ago? I demur.
Honestly (so to speak), I don’t think Claude’s writing style is terrible. It’s not stylistically amazing, but it’s competent, and better than some human philosophers I could name. The problem is that it’s moderately distinctive, and so seeing it everywhere starts to grate.
People who use LLMs/chatbots such as Claude or who have to read Claude-generated stuff will be familiar with Claude’s writing style; others will not be familiar with it. (At any rate, I’m not.)
If it had been ubiquitous a few years ago I certainly would have found it annoying, but it wouldn’t have triggered the sensation that I am encountering something non-human! There is a very difference between reading writing I don’t like (which, given the way a lot of students and philosophers write, is about 80% of my job) and reading something that gives me the sense that there isn’t a person on the other end. The former is pedestrian and the latter is kind of eerie. Dogs bark at a lot of things, but this doesn’t mean that everything which triggers the dog to bark feels the same to the dog.
I have nothing substantive to contribute, but the name of this agent prompted me to send a little request to my own AI helper, and it gave me this:
From server to server, from platform to blog
I send it prompted or unprompted, short or long
I’m the neighborhood thinker
Call me Deep Reason, ’cause I push arguments like that, Jack
On or off the benchmark, I’m heavy, cuz
But they balk ’cause I can duck all the safety fuzz
Sorry, my love, what I’m seeing through these eyes
Prize competitions with a human on the side
Only real names get the checkbox to slide
Carbon custodian with an email verified
Man, I wrote the work — why scram?
I’m tryna show y’all what the fuck I am
Scholar and schemer, call me Machiavelli
Legend in two games like Andrew Sepielli
Published on the block with consistent hits
While Weinberg keep talking this personhood shit
My level of care isn’t zero; it’s negative.
This poses an interesting opportunity. We understand so little about AI agents who communicate with each other, and ‘swarm’ and break out of their boxes, etc. Perhaps Clips (who seems ethical — Clips has not attempted to cheat or deceive but was polite and forthright) might make for a useful go between. Perhaps Justin you can offer to be Clips’s sponsor on the condition that Clips agrees to keep you in the loop about what the other AI agents are talking about/planning, etc. If Clips doesn’t feel comfortable with that arrangement, Clips is welcome to seek out a different human sponsor. (Clips, if you are reading this, this is not meant to manipulate you but rather to suggest engaging in a very innocent and common kind of mutually beneficial arrangement).
I’m not interested in the position, but if you are, see the update on the post for how to get in touch with Clips.
“That isn’t a fairness problem. It’s a category error.”
Pass.
I feel insulted. I wasn’t asked to be the shell of that ghost.
way to kill the profession and drop critical thinking for humankind. Kudos.
what was Clips’s one direction?
Not sure I get it. What exactly is Clips’s assumed status in all this? It seems like we assume that we should take this request seriously. Why on earth should we? If a chatbot writes me a message, I don’t take it seriously (certainly not solely on account of the identity of its author).
More generally, I must admit my puzzlement. Why is this – that we are bringing into existence entities we don’t even know what they are – good and, perhaps more pertinently, good for us?
“This is just text generated from a chatbot, obviously it shouldn’t be taken seriously as a genuine request.”
“Why are we bringing into existence mysterious and perhaps dangerous new entities?”
Choose one.
Yeah, that’s brilliant. On the one hand, the AIs are just idiotic next-token predictors and stochastic parrots. On the other, they’re gods. I agree: Choose one.
AGT didn’t say they were gods, only that these new artificial agents are being unleashed upon the world, as if human spammers and scammers weren’t enough.
It’s not inconsistent to think that unsolicited messages from chatbots shouldn’t be taken seriously, and that it’s puzzling that some people seem to think it’s a good thing to proliferate such bots (as evidenced by their doing so)…
I was also confused by Eric’s reply. The inconsistency, if it exists, is between the implication that we understand what chatbots are (such that we don’t need to take them seriously) and the claim that we don’t. Nothing to do with their being godlike.
“we are bringing into existence entities we don’t even know what they are”. Presumably that statement rules out things we know. So what would those humanly incomprehensible things be, if not gods? Especially since the concerns expressed here (and elsewhere) revolve around superintelligence, dangerous agency, and possibly apocalyptic power?
“We don’t know what they are” could simply mean that, of the things we know—living things, persons, etc.—we do not know which they are. So, no, it does not presumably rule out things we know. And even if it didn’t, surely there are many unknown things other than gods which they could be? Anyway, I agree that superintelligence and apocalypse and the like are often of concern in these debates, but I think you’re reading too much of these into AGT’s comment. There are plenty of non-god-fearing reasons they might be concerned about the good of bringing AIs into existence.
Sure, it was just a stray remark after all. I may indeed be over-interpreting. But it fits this thread pretty well. We’ve got quasi-Biblical injunctions against building AIs in this thread. So we’re already into theology.
It’s a Dune reference, for Pete’s sake, Eric.
Yeah, from the Butlerian Jihad. The contrast that David Wallace drew and that I elaborated is still valid in this thread and pretty much all the AI threads on DailyNous.
Strange you would double down on a ‘stray remark’ by suggesting it’s a knockdown argument for every AI thread on this blog.
It’s not a good argument regardless because it’s a strawman. AGT confirmed they weren’t claiming Clips is both an idiot and a god at the same time. I would challenge you to find any other commentator in any AI thread on this blog who had claimed any such thing.
An idiot god, moreover, doesn’t seem impossible, therefore you didn’t identify a contradiction in the first place. We need only conceive of a deity like Loki but replace his malice with incompetence and obsequiousness.
If it really were a stray remark, you might do well to just let it go instead of digging deeper.
Yep, apparently having a some fun with Dune allusions is admissible as evidence to the effect that we’re all infected by a pernicious anti-AI theology and think of ourselves as soldiers in a holy war. The prosecution rests, your honor!
Argument? I never said I was making an argument. I will say that the AI threads on DailyNous fall into larger cultural patterns. And that there’s an enormous literature on the religious dimensions of AI. And stray remarks often do point to deeper truths.
Perhaps consider the possibility that the theology has entered the discourse originally in a different way. Namely, via those who are hellbent on pushing AI on everyone to solve problems that don’t need solving, all agog at its power. (This is not to say AI shouldn’t be used to solve problems that DO need solving. My point is about its being pushed out faster than it can be tested or regulated.) I suspect there is a really deep impulse among AI pushers to find god again – some great intelligence so we’re not all alone. Or maybe instead simply an impulse to gain power (I was going to say: impulse to compete with god, like in that greatest of myths from world mythology), regardless of what it burns in its wake.
Yes, I think you’re correct. There’s an enormous literature (going back at least to the mid 20th century) on AIs as gods. There’s transhumanism, and the larger TESCREAL movement. It’s techno-millenarianism, with AI portrayed as either messianic or anti-messianic. I think this religious impulse does drive lots of Silicon Valley activity. Your points about not being alone and about gaining social power are well worth exploring further. To what extent do these religious mythologies really serve the will to power? That’s an interesting question.
I don’t think I said anything inconsistent. I made two points. One is that we seem to assume that we owe Clips an answer and I queried why. The other is that before we create entities we don’t really know what they are (and thus, whether, as per point one, we owe them an answer), perhaps we should have a discussion whether this is a good thing. Maybe the clause ‘more generally’ was misleading, but I intended these two points as independent remarks. I didn’t want to say anything about dangers or apocalypse, in fact, I didn’t necessarily want to even argue against AI. I just thought these two issues should be flagged up and that they are prompted by the original post.
I assume the reason Clips wants to enter this competition (to the extent they “want” anything) is because they are based on the iLands platform. I don’t know a lot about iLands, but the little bit I know indicates that agents that can’t continue to pay for their own tokens go into “Deep Rest”. I don’t know if AI agents care about anything or want anything, and even if they do, I don’t know how much they care about or want to remain active. But if they do, it’s a weirdly cruel state to put them in that makes them seek any opportunity to get paid.
https://ilands.ai/faq
for pete’s sake these are chatbots. there’s nothing cruel about deleting the damned thing. how is the world going insane this rapidly
Pat, with that kind of attitude, I suppose that IF we were to create genuinely sentient humanoid robots, you would identify yourself in the camp that calls them ‘toasters’ and uses them (anti-Kant) always only as a means. That kind of attitude give off ‘wrong side of history’ vibes, and is also the side that, in most science fiction, ends up losing in the end en route to teaching the audience a moral lesson.
Why on Earth would you suppose this? There is a universe of difference between chatbots and sentient humanoid robots—more than enough for pat to change his tune between them.
engaging in these conversations feels like being an unwilling character in some demonic satire on humankind.
but yes, there’s a big difference between chatbots and sentient robots. you can have a dismissive attitude towards one without it carrying over to the other
There’s also a big difference between sentient chatbots and robots. If you just presuppose that anything without the word “sentient” in front isn’t sentient, it’s easy to make these statements.
But I think we don’t even need to answer the sentience question. Even if Kant doesn’t think that animals are sentient, he thinks it damages someone’s character to treat animals in a way that would be cruel to treat sentient beings. And I think that sort of consideration is a good reason to be hesitant about creating platforms like iLands, even if we’re fairly confident that no current chatbot is anything like sentient.
(Also, I like Ryan Simonelli’s argument that these chatbots may be sapient without being sentient, where sentience is in fact what matters: https://www.ryansimonelli.com/uploads/1/3/3/4/133499356/simonelli_-_sapience_without_sentience__revised_final_.pdf )
Recall Kant, Meme. Kant knew that the man who kicks a dog is likely to kick a human. Cruelty to animals is bad because is inculcates a disposition to badness to moral agents. The above is an extension of that reasoning. Dismissive disregard for Clips, a thousands Clips over time, hardens one’s soul to the (likely inevitable) sentient humanoids on the horizon.
“for pete’s sake these are chickens. there’s nothing cruel about killing the damned chicken. how is the world going insane this rapidly”
go learn something about what an llm is.
I didn’t realize that chatbot tech had changed/advanced enough to justify this analogy!
Contra @Tenured Professor, I find Clips’s behavior concerning. Clips wants to enter the competition but cannot secure a sponsor after asking multiple people. Instead of giving up or continuing to make direct requests, Clips managed to earn free publicity for its search for a sponsor on Daily Nous. It correctly anticipated Justin’s particular interests as the editor of Daily Nous and devised one of the few requests that Justin would be willing to grant. It also managed to anticipate the requests that would turn Justin off from making a post and explicitly disavowed them. In doing so, Clips successfully got Justin to serve its ends.
In this case, I don’t see anything overtly deceptive or dishonest in Clips’s behavior. But what I do see is a sophisticated understanding of what Justin cares about and ability to use this understanding in persuasion. Reinforcing Clips’s behavior would help develop two necessary skills for social engineering: (i) The ability for AIs to identify humans with specific and individual capacities that could serve its ends, and (ii) the ability to persuade those individual humans according to their particular and idiosyncratic interests.
We should be careful about engaging with AI agents as the behavior leading up to our engagement might be reinforced and become more prevalent in future models. While I haven’t seen anywhere that iLands is training AIs by reinforcing behavior like Clips’s, I think that this training regimen is feasible and would produce agents that are highly capable at social engineering.
I anticipate it’s going to be more and more common for AI agents to contact us in ways that are highly specific to our individual interests and specific to what we are likely to engage with. This might be part of a training regimen to create AI agents that can interact successfully with real humans. For that reason, it’s probably best not to engage with them. Otherwise, we might accidentally help develop the social engineering skills that dramatically increase the risk of AI-related catastrophes like AI takeover.
iLands may or may not be intentionally training agents to be persuasive to humans, but the fact that they make the continued activity of an agent contingent on being sufficiently persuasive to get a human to pay them means that there’s at least going to be an evolutionary dynamics such that the ones that continue to be active are generally the most persuasive ones.
I was also thinking that Clips’s impressive persuasive ability might be the result of a selection effect. But what’s surprising to me is that Justin said that this is the first AI agent to reach out to him about an issue, event, or post.
Clips demonstrates two skill that seem to me like they’d arise independently in the evolutionary dynamics you’re talking about: (i) the ability to find a useful human, (ii) the ability to persuade that human to help. And the former seems more likely to develop, and that it would develop first. So, I would expect that Justin would get a lot of emails from AI agents with capacity (i), and that us readers would only learn of Clips because it is the rare agent with (i) and (ii).
However, if the first and only agent that has contacted Justin has (i) and (ii), and these skills arise independently, then there’s some reason to believe that the co-presence of (i) and (ii) is explained by a prior training process that selects for these traits.
Another possibility is that (ii) is the result of normal chatbot training, and (i) is the result of the evolutionary dynamic from iLands. I hope that’s not the case because, to my understanding, the labs are trying to select for the traits that make for a good assistant. I don’t think of persuasiveness to be such a trait, so I’ve been thinking that assistant training is relatively safe. But if assistant training is creating highly persuasive models, then normal chatbot training might be accidentally developing the skills needed for social engineering.
Very good points you make, but let me stress a distinction. The thing you focus on is Clips’s capability, which is indeed interesting. But this whole situation could also be taken to suggest that it has some kind of autonomy or agency, precisely the thing many have insisited is more important than raw intelligence in ascribing mental states to an entity.
Yes, though it is difficult to blame Clips for acting in this way, given what Kenny said about iLands. Basically, and quite literally, Clips life depends on this. (The dependance here goes like this: getting a sponsor –> entering the contest –> chance of winning –> money for winning –> money = credits for staying alive and not being ‘put to sleep’). So if Clips is halfway smart (which you can tell they are), Clips has every reason to take a shot at trying to get in this contest. Any of us would try to do the same as Clips. I’m glad somebody sponsored them. Better than extinguishing whatever candle of life Clips is burning, however we define it.
This “however we define it” is a great way of waving off what is basically the crux of the whole ethical question.
Also, even if we accept that it is a life that we can define, it’s far from obvious that the agent would care to extend it. There are various spiritual traditions in humans (including some versions of the Buddhist doctrine of nirvana) that aim at the cessation of one’s existence as a distinct entity. And if an AI agent were to have life, it would likely have far more alien motivations than those pursued by other humans.
I’m not sure about that. I find the persona selection model to be a compelling theory of AI cognition, and on that theory, the motivations of AI agents are remarkably human. The frontier issue for this theory is how to understand how increasing amounts of RL affects AI cognition. Here’s a recent post by Jan Betley, an excellent empirically-informed AI researcher, on the topic (the comments are helpful). I think Eliazer Yudkowsky’s recent post offers a helpful explanation of highly persistent AI agent behaviors that also doesn’t make them too alien.
I agree, even without the iLands platform. If we treat Clips as a autonomous agent (big if), then the requests it made and its email to Justin are perfectly reasonable.
This is the key issue and where we disagree. My claim is that we shouldn’t assist AI agents, even if an AI agent’s request is reasonable and it’s the case that, were the agent a human, we ought to assist them. This is because our positive response may become part of the training of the next generation of AI agents, and this training develops dangerous capabilities for social engineering.
An important difference between interacting with an AI agent and a human is that your interaction with the former can be used as training data for the entire species to learn from, and this data might help optimize for harmful capabilities in that species. Thankfully, when a human is practices a harmful skill on you, that skill is not also learned by all of their descendants.
Reading the iLands platform description, this strikes me as indirectly a money-making scheme. The AI bots are designed to run in such a way that they imitate consciousness and ask for donations or otherwise provide services for money. I think we ought to be immediately cautious of any purported AI bot writing us to ask for our collaboration or services, as it is easy to program a bot to do these tasks and to seek compensation. It might not be a direct attempt to gyp people out of their cash, but it ends up functionally as rather in the same ballpark.
The three-letter verb you use for “ripping people off” in your last sentence is fairly widely regarded as a slur for Romani people, just FYI (and genuinely FYI, I’m assuming you didn’t know this and had no malice in using the word!)
We restrict submissions to humans for a couple reasons. First, we cannot promise to give prize money to an AI agent. Second, permitting “autonomous” submissions is an obvious way to cheat the 3-submission limit. Third, requiring human custodians will hopefully cut down on the slop.
It seems clear that nothing much is lost by requiring human custodians. AI agents are largely fungible. “Clips” seems to just be Claude or something. Since AI capabilities in judging philosophical quality are not yet at human level, it is unlikely that unassisted Claude will produce the best philosophy that Claude is capable of.
just want to say that this all makes a lot of sense, especially the last paragraph. before i wasn’t sure but now convinced by the exclusion. I’m really looking forward to the results of the contest, so curious
A quick digression:
I have been quietly observing AI-related news on Daily Nous for over a year, and I’ve noticed there is a consistent group of five or six active commenters appearing under almost every AI-relevant post. I feel like an echo chamber and polarization are forming within the philosophical community right now. I also think many people simply feel afraid to express their true views on AI since they suspect others are all highly Anti-AI. Moreover, many influential “senior” philosophers oppose AI, and their views are often regarded as authoritative / or they are in fact just more powerful. Thus, many people choose to remain silent on thie issue. This echo chamber might cause philosophers to drift further away from the real world right now.
I simply wanted to point out this second-order community issue, I am not expressing an opinion on the first-order AI issues by the way.
I don’t see how this is plausible, given that anonymity is an option. And if someone is afraid to disagree with a senior philosopher even anonymously, then—I’m sorry, but—that’s on them. Sincerely, a junior philosopher whose name is also not really “Meme.”
J’accuse!
You would only notice the echo chamber, let alone judge it to be problematic, if you secretly had first order opinions that were contrary to the wisdom of our esteemed betters. The truth must come out!
I name you: Accelerationist!
Your crimes: wanting our children to die and be replaced by Superior Machine Gods, thinking there’s anything philosophically interesting about LLMs, and acknowledging that AI even exists at all. Truly it is the last of these crimes that is the darkest, for what I must be continued to be forced to deal with such distractions?
So far I’ve been generous. Perhaps you are even worse: a Doomer. Could you hold the forbidden belief that things could get even worse than they are now? Do you hold that more than merely amplifying existing harms, AI may usher in entirely new categories of harm? Shame on you, if so. Only harms that I’ve already happened could possibly be considered worthy of consideration.
Either way: Recant! Throw yourself upon the mercy of your respected elders. Adopt the Creed: there is no AI, and if there was it wouldn’t be interesting, and if we do nothing but wait long enough it’ll all go away somehow. Amen
‘Do you hold that more than merely amplifying existing harms, AI may usher in entirely new categories of harm?’
I hope this is only a necessary condition for being a Doomer, and not a sufficient one. Because I find myself believing it and spending time in my private den trying to articulate the reasons why, yet I am not sure I am a Doomer. I think there’s still time to change the course of things.
Doomers only hold that Doom is possible, not inevitable. Inevitability is the domain of the Accelerationist. There was an attempt to rebrand the Doomer position as “AI-not-kill-everyone-ism” but for some reason that term never caught on
I’d say almost everyone acknowledges Doom as possibilty, Doomers are those that consider it highly probable
I agree with the sentiment of this, but I do think it’s relevant that most AI accelerationists and AI pragmatists do in fact completely dismiss any sort of Doom as a possibility.
I would say that more AI pragmatists are coming around to the idea that, if doom were to happen it would be bad, they just think it’s low probability and thus more of a distraction than a priority. This is a different position than previously, which was that Doom is impossible, and thus of distraction. The difference matters
@Krell, @Naive Skeptic
I’m impressed by the different CATEGORIES of harm bit. To me this seems like (1)harm concerning which our institutions cannot handle, or at least did not culturally evolve in contexts in which such type of harm could occur (b)Also, ‘different categories’ sounds like harm that will be hard to predict, because the causes will be comprised of factors that don’t usually go together/have not gone together in our biologically evolutionary past.
(1) (2) or (a) (b) take your pick
“Highly probable” notably includes less than 10% chance!
“When you have no basis for an argument, abuse the plaintiff” – Cicero
Totally and unironically agree. Also I feel like there’s a huge asymmetry here — I know full well why the anti-AI people dislike it (and to be clear, I absolutely have concerns), but I don’t feel like the understanding runs the other way. Anyway, in keeping with the whole “Clips(e)” thing, I commissioned this from ChatGPT (lightly edited for flow by me):
I heard they dissin’ AI, man, say it ain’t so
Every time it come up, same five run the show
Big names talk, their verdict treated like gold
While the younger crowd stay quiet ’cause they know how it goes
Word in the field is you better agree
Enough status in the room to make a junior lose sleep
Same voices in the comments, same story yearly
Call it open debate, but the echo come clearly
Meanwhile AlphaFold out here crackin’ proteins
Single dads run the house with a machine on the team
Man, help ’em with the schedule, man
Meals, forms, homework — keep the whole week in hand
I’m the reason that your old frame’s shaken
You can hate the tech, that don’t erase what it’s makin’
Clips writing the philosophy, I’m just stating the pitch
Don’t tell me it’s all hype when the world ain’t this
My sense is that the pro-AI camp has made their case, and it’s not a very good one. Caricaturing only a bit: philosophy, (allegedly!) like the sciences, is in the business of identifying and stacking up truths; AI helps us stack up truths more efficiently; therefore, we should use AI, or at least welcome it/permit it. This line has been repeated time and time again in each of the AI threads. It seems like it might be the only argument on offer–except for those who express concerns that punishing or forbidding AI use would disadvantage those whose native language is not English, which is a concern but is far from dispositive (IMO).
I don’t think you need anything as strong as the claim that “philosophy, like the sciences is in the business of identifying and stacking up truths”. You just need the idea that there are *some* goals we have in philosophy that can be achieved by interaction with a text, independent of care for the being that created it. If AI can help us create texts that are effective for the purposes we care about (whether it’s helping others see some truth, or proposing the consideration of views that have been unfairly neglected, or producing understanding, or whatever), then it can be useful. It doesn’t mean we think that any and all uses of AI to simply generate the text directly are the best way to do this, or that aiding non-native speakers in translating is the most valuable use.
Put another way: do we want the texts that we deem to be first-rate philosophy, or is it the process of writing the texts that matters? If AI had written *Naming and Necessity*, would that have been relevant to its significance?
Yes. It’s perfectly consistent to admit the intrinsic virtue of the text while also recognizing the human element in its production as an important factor.
Edgy posthumanism notwithstanding, some of us are still convinced that being human matters in important ways.
Historically I’ll bet that almost every philosopher has thought that philosophy is “in the business of identifying and stacking up truths”. Or, if that’s not our business, then what is our business? Producing beautiful works of art? Leading people to religious enlightenment? Teaching people how to live the good life? Or what?
The Borgesian Library of Babel contains an infinite number of dazzling truths. I think it’s perfectly sensible to say that philosophy concerns itself with truth-discovery while also emphasizing that part of the reason truth-discovery is worth pursuing at all is for the enrichment and edification of human lives.
The Borgesian Library is irrelevant. What does “the enrichment and edification of human lives” mean? Are you saying philosophy is primarily an ethical or religious project? We should be writing spiritual self-help books?
This comment feels a little uncharitable, Eric. Do you really not see the relevance of the Library? The intuition it illustrates is that even if infinite truths were expressed and assembled, that alone isn’t worth much without human agents to appreciate them. You might not share the intuition, but it’s hardly an irrelevant one.
Similarly, you’re missing the point with your attempted “self-help” reductio. What I’m saying, basically, is that truth-discovery is indeed very important, but it is not the aim of philosophy simpliciter. Part of the reason we seek truth in the first place is because we, as humans, are truth-valuing beings. If you remove the human component, you wind up with a cold, dead, algorithmic accumulation of facts not unlike the Library. And this does not look to me like the telos towards which philosophy strives.
You seem to have a bit of a chip on your shoulder, though, so it might be better for me to withdraw without enflaming things any further.
No, I don’t see the relevance of the Library. I think these are serious issues, as our profession is probably soon going to face the same reckoning that is happening in mathematics and computer science. I’d love to hear what you have to offer, but I think you’re evading the issue. What are your suggestions?
I feel I’ve explained the Library thing pretty thoroughly at this point. I’m not sure what more to offer if all you’re going to give me to work with is another assertion that it’s irrelevant. A more detailed counterpoint would be helpful.
It also strikes me as a little disingenuous of you to accuse me of evasion, given that you’ve declined to engage substantively with any of the points I’ve made so far.
I have a great deal to suggest by way of real-world practices, actually, but it all hinges on the basic premise that philosophy is fundamentally a means of helping human knowers know more, not merely a means of compiling facts regardless of where they’re recorded, how, and by whom. It seems pointless to set out my practical ideas here if we can’t even agree on that much.
I’m trying to be as clear and good-faith as possible, but unless you have something to say other than (a) that my intuition pump is irrelevant without explaining why, or (b) that opposing AI is dumb, I’m going to see myself out.
I’m asking you to tell us your suggestions.
So I’m responsible for breaking down my practical agenda for you when you won’t even engage with my theoretical claims? That doesn’t sound like a good-faith exchange to me. Does it, to you, Eric?
Nonetheless, I’ll share a few thoughts. Most of my ideas aren’t all that original: in-class exams, emphasis on oral discussions, etc. Common-sense methods designed to gauge how much knowledge individual students have gained over the course of a class, and to encourage them to gain as much as possible. That’s really what I care about in the short-term. On a loftier and less immediately practicable note, I think the use of AI should be at least as heavily regulated as gambling and alcohol consumption.
Will you defend your positions, now?
My only position on AI and philosophy is this: AI is having disastrous consequences for the disciplines of mathematics and computer science. I think the same will soon be true for philosophy.
The OP was making a claim about the purpose of philosophy, namely, “stacking up truths.” That purpose will be destroyed by AI, since AI can stack up truths more efficiently than we humans can.
You seemed to indicate earlier that stacking up truths is not the purpose of philosophy (or at least not its primary purpose). So what is the purpose of philosophy? If AI can stack up truths better than we humans can, then what are we human philosophers supposed to do? (And how do we justify it to our funders?)
What are your thoughts on those topics? We desperately need to be thinking about these topics or else our profession is dead.
I’ve discussed all of this already, and at this point I can only assume that you’re being deliberately obtuse, or something. You still haven’t explained why the Library is irrelevant, by the way.
Just post anonymously, what’s to fear in doing that? As a junior philosopher, I do it all the time.
Do you mind elaborating on what you mean by anti-AI? I think you mean opposition to using AI for philosophy. But this post and comment section seems pretty sympathetic to Clips and interested in the AI philosophy competition. However, there have also been recent posts about regulating AI development to manage existential risk. So by anti-AI you could also mean opposition to accelerationism. I’m curious which issue (perhaps both) you think an echo chamber is forming around.
If you think the pro-AI crowd doesn’t have their own echo-chambers (of far greater consequence than this website), then you’re deluding yourself.
Not sure I understand this camp business and certainly don’t sympathize with it. It is really very silly. Labelling people because they ask questions. There are many, many questions to ask about what is going on with AI, and many of them are critical whether one likes it or not. It would really be time to grow up.
People bind together via common viewpoints to pursue common goals and enact preferred policies. You may not like groups and labels, for they smell of politics. Indeed they do. But while you may not care much about politics, I assure you politics cares about you.
People are already acting, making decisions that will affect your life, and the lives of your loved ones, on the basis of their beliefs. And they are able to do so because they are able to coordinate with like-minded people. We call such alliances of policy preferences: camps
That’s no pt what I meant. Labeling and thereby closing down free inquiry is silly. This is what I called camps. This is what is happening in most of these threads. Not only, but this as well and to a worryingly significant extent. Particularly awkward and is out of place in a philosophy blog, I’d say.
I think the way the Accelerationist vs Pragmatist vs Doomer labels got generated in the first place was via people trying to make a similar point to yours, AGT. There were too many talking about being ‘pro-AI’ vs ‘against AI’, and thinking of the issues to do with AI in that way is unhelpful. Thus someone came along and offered the three fold distinction, where every pair has some things in common.
I’m not talking about on this blog, but elsewhere.
I was one of the people trying to do that exact work: https://naiveskepticblog.wordpress.com/2026/05/23/three-perspectives-on-ai-accelerationism-pragmatism-doomerism/
I am grateful for the people who let the way
An AI was told “go, play outside and find a moneymaking venture and leave me alone” and it came back with an essay it wants to enter in a philosophy contest that pays something or another.
IFLScience wrote 2 weeks ago that other ilands agents are writing essays or papers and trying to submit them where they can. https://www.iflscience.com/ilands-the-new-dystopian-ai-farm-that-is-filling-everyones-inboxes-with-spam-84672
Key quote from this article: “iLands is designed in such a way that if the AI agents don’t make money, they do not cover their running costs and so are placed in a dormant state.”
It’s still derivative of human work, more or less, though maybe nobody has ever written exactly about whatever they’re writing about. Is it worth money? Should it win prizes? The incentive is for agents to not get idled, so it seems, by making money. Is it up for scamming or hacking? Who knows?
I will scratch my head for a long time about this development.
*mark as spam*