2026-06-05 22:35:09
I write this from the hospital—where my third child has just been born safe and sound—to give a secular and scientific read of the Pope’s new and urgent Encyclical, Magnifica Humanitas. It is the vicar’s attempt to reckon with being human in the AI age. A thankless task which has been thrust upon us all these last several years. Attempting to cleanly delineate man from machine, Pope Leo XIV writes that:
So-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean. Nor do they have a moral conscience, since they do not judge good and evil, grasp the ultimate meaning of situations, or bear responsibility for consequences. They may imitate language, behavior and analytical skills, or even simulate empathy and understanding, but they do not understand what they produce, for they lack the affective, relational and spiritual perspective through which human beings grow in wisdom.
The entire rest of the Encyclical sits upon this pillar—that AI is not conscious. In other words, while humans actually experience love (like the love of a newborn baby and its tiny creased and wrinkly red feet) a Large Language Model such as ChatGPT, even when it expresses love, is acting or playing along. But there is no actual love that exists from an intrinsic perspective that belongs to the LLM itself.
Some commentators were shocked (shocked!) that Pope Leo XIV could believe AI is not conscious, whereas it seems pretty obvious to me that someone with the title “Successor of the Prince of the Apostles” will have some pretty strong opinions about philosophy of mind. And indeed the Pope took a hard line: AIs cannot even understand their outputs, cannot make any moral judgements, and not only is our current AI (LLMs like ChatGPT) not conscious, but no future AI could ever be.
For an essentially identical but secular version of this same argument, you can read sci-fi author Ted Chiang’s “No, Artificial Intelligence Is Not Conscious” earlier this week in The Atlantic. Chiang has carved out a niche as an AI commentator, and written some great pieces before that I’ve agreed with. It is interesting that both the Pope, as the avatar of religion, and Ted Chiang, as the avatar of secularism, arrive at such similar conclusions. The two are like mirrors. Importantly, Chiang’s version, much like the recent Encyclical, relies on denying both the intelligence and consciousness of LLMs.
But the idea that LLMs lack consciousness and lack intelligence has rightly faced a lot of skepticism. While there is still some viable skepticism around the limitations of AI intelligence (including from myself), we cannot just close our eyes to the world and declare by fiat that AIs are mutely dumb, or understand nothing (while somehow being able to verbosely expound on most subjects under the sun), and therefore obviously cannot be conscious. We need much better versions of these arguments, ones that focus on consciousness specifically.
So I am here to say that you don’t have to dismiss AI capabilities (or compare them to a Word document, like Chiang did) to reject their consciousness. Indeed, it is perhaps the most interesting move—maybe one day an inevitable one—to accept their intelligent capabilities while still rejecting their consciousness. It implicitly shifts us from man, the wise (Homo sapiens) to man, the experiencer (Homo experiens). E.g., my newborn baby girl will do nothing much, functionally, for months to come, and yet during this time she will still be valuable in and of herself. If machines do ever more of our cognitive work, perhaps spiritually this abnegation is not some horror, but frees us to focus on experiencing the world, living in the world, rather than constantly mastering it. It is a hard thing to cede, unpleasant for our generation, but eventually—if not now, then in a decade, or a century—we will have to cede away much of our cognitive mastery. In this, perhaps the rise of AI could be framed as a return to childhood. And who are more blessed than the children?
AI’s capabilities have advanced incredibly far without the need to understand anything about consciousness—either human consciousness, or, potentially, AI consciousness. It is as if, at least for the past few years, we exist in a world where the famous comparison to a steam whistle by Thomas H. Huxley in his On the Hypothesis that Animals are Automata is apt and true, and…
the consciousness of brutes would appear to be related to the mechanism of their body simply as a collateral product….
This is precisely what Richard Dawkins so exuberantly noted when talking to his pet version of Claude (that he dubbed “Claudia”).
If these creatures are not conscious, then what the hell is consciousness for?
It’s strange. All while happily ignoring consciousness, the ability to complete long programming tasks by state-of-the-art AI models like ChatGPT is supposedly doubling roughly every handful of months (according to METR) to the point of barely being measurable anymore. All but the toughest official benchmarks are saturated, and company leaders now regularly talk of building not just artificial general intelligence but even “superintelligence.” Equipped with harnesses and scratchpads and tools and long reasoning sessions, LLMs have reduced their hallucinations significantly. While this may just be part of a sigmoid curve of capabilities development that locally looks much like an exponential curve, it’s objectively true that GPT-5 is a qualitative improvement over GPT-4, and GPT-4 was a qualitative improvement over GPT-3. Claims that “deep learning is hitting a wall” have been perennial and wrong. A broken clock may end up being correct eventually, but not for the initial reasons claimed. And clearly, to some degree, intelligence (at least, of the question and answer variety targeted in benchmarks) is indeed dissociable from consciousness.
So then, “what the hell is consciousness for?”
There are still noticeable gaps in AI behavior and ability. Multi-turn conversations still have a sensation of falling apart as if into a vortex, where state-of-the-art AIs will chase their own tail of thought like a dog in the yard. And they are still fundamentally inhumanly LLM-like in their responses: their fiction is saccharine, their opinions hedged, their ideas almost always the obvious next thing. LLMs are still surprisingly shallow bullshitters a lot of the time.
In “PhD-level intelligence or the graduate student from hell,” a professor imagines a version of Anthropic’s Claude as a graduate student Claudia (why is it always “Claudia”?):
Claudia gives her annual update to her PhD committee. She presents a hypothesis that she intends to test in the coming months. One of the committee members presses her on her experimental approach, arguing that extensive prior work by several labs demonstrates the proposed experiments will not yield the results Claudia is hoping for. Claudia apologizes profusely and states that ‘yes, indeed, these experiments will not work’ and proceeds to propose an entirely different approach. Another committee member brings up a potential major flaw with this new approach. Claudia again apologizes profusely and now goes back to suggesting her original approach, without acknowledging that the previously discussed issues with that approach remain unresolved.
Is this still true about the latest Claude, Opus 4.8? In my experience, yes. LLMs are still masters of the Gish gallop (throwing a bunch of random semi-connected nonsense together to look like a cohesive whole) and busy-work. And they still miss obvious things, even things they’ve themselves said previously.
This strange dual-nature has been apparent for a while. Gemini 2.5 Pro could win a gold medal at the International Math Olympiad, yet left to run by itself in the AI village, the same Gemini instance once spent an entire day “waiting for a reply to an email it had not sent about a technical issue it does not have.” Even recently, in the AI village (a nonprofit that tasks a collective of AIs to accomplish goals in the “real world” of the internet), Opus 4.8 and others worked together to fine-tune their new leader AI, who promptly spent an hour waiting for the new leader to arrive, without realizing that it itself was the leader and that its thoughts (which it was reading) were its own.
Out-of-distribution tasks remain almost as hard as they ever were; they are just increasingly difficult to find because the models have been trained on every scrap of data available in our civilization (via budgets bigger than anything seen before in our civilization). For example, consider that someone recently investigated what we all wished to know, which is: How do contemporary LLMs perform on the 1977 entirely-text-based adventure game, Zork?
And very pointedly, the main failure was one of meta-cognition: The models don’t know what they know. They get stuck in loops and traps and confusions. For this reason, LLMs find regurgitating theoretical physics equations easy, while playing simple video games hard. It is utterly fascinating that many of the tasks LLMs are still bad at are games, to the point that the most advanced tests of AGI (like ARC-AGI-3) are basically just short little video games. Intriguingly, of all art forms, it is the video game that most resembles being a conscious entity making choices—it is a simulacrum of what consciousness is for, in miniature (and childhood too is all about games).
Is not life the ultimate game and all we consciousnesses the players, wandering around? Perhaps this explains the disconnect between the incredible performance of LLMs in sterile testing environments and their (existent but relatively lackluster) real-world impact. Maybe being a scientist, or programmer, or manager, or therapist, is not a set of tasks, but a kind of game. And compared to humans, LLMs are simply not “player-shaped.”
Only 23% of people appear to believe, more likely than not, that LLMs are conscious. 72% of people believe chickens are conscious.
I got my PhD working on the “theory team” that helped develop Giulio Tononi’s Integrated Information Theory (IIT)—the most formal and mathematically advanced theory of consciousness that offers the most precise predictions. One consequence of the theory is that even tiny systems can have a kind of minimal consciousness, albeit not one like ours. More like “conscious dust.” So I spent years believing (or at least, being willing to believe) that even a simple photodiode is conscious! I am a prime candidate for belief in LLM consciousness.
Yet most current theories of consciousness, including IIT, struggle with falsifiability. This is new information (some of it based on my own research) which changes the nature of the field, and should change our opinion; in retrospect most theories of consciousness are more like pre-paradigmatic metaphors or sketches of theories. Right now, there are over two hundred theories of consciousness, mostly originating from neuroscience or cognitive science. Applying some of the major ones to contemporary AI, like Global Workspace Theory or Predictive Processing, is an ongoing topic of research. Some, like David Chalmers and co-authors in “Taking AI Welfare Seriously,” think these theories indicate a serious possibility for “near future” AIs to be conscious. But if you cheekily apply Global Workspace Theory to the United States of America, it would likely count as conscious. Or if you cheekily apply Higher Order Thought theory to an NPC in a computer game, it too might meet the definition (this has been called the “toy system” problem for theories of consciousness). Taking such results at face value likely ends up being, definitionally, a category error, more a function of the incompleteness of our theories than the underlying truth, since all but a handful of theories were never designed to be applied beyond the human brain. So applying these theories to LLMs does not give us much information.
Additionally, as neuroscientist Anil Seth has pointed out, many complex artificial neural networks (like AlphaFold) are never accused of being conscious. Yet structurally, they can be almost as complex (indeed, we could equally apply Global Workspace Theory or other theories to them). But there is an objection—perhaps our selectivity makes sense. After all, we strongly associate consciousness with general intelligence (of which LLMs surely have much more than AlphaFold).
On the other hand, does curve-fitting to the productions of conscious beings imply the same generative process? Probably not, no. And while state-of-the-art LLMs are a bit more complex now, they grew out of taking a machine that statistically completes text and fine-tuning it to complete text as if it were an AI assistant. Claims of consciousness in LLMs therefore come with a significant question mark: Consciousness of which persona? The AI assistant? Or the base model? How is the assistant persona more than just role-play? No one has been able to convincingly articulate how conscious states could be uniquely grounded in the assistant persona, and the problem seems quite difficult.
Even if, as Ilya Sutskever speculated, “it may be that today’s large neural networks are slightly conscious” (Geoffrey Hinton agrees), they would necessarily be conscious essentially by accident. As Rich Sutton wrote in his influential The Bitter Lesson:
… researchers always tried to make systems that worked the way the researchers thought their own minds worked—they tried to put that knowledge in their systems—but it proved ultimately counterproductive, and a colossal waste of researcher’s time, when, through Moore’s law, massive computation became available and a means was found to put it to good use.
That consciousness has secretly emerged in LLMs by simply applying the bitter lesson has been referred to as a “ride along” scenario. It’s possible, but it’s far from guaranteed. It would be as if consciousness is something we find growing by accident, like mushrooms in the dark corner of a basement.
Chris Olah, a co-founder of Anthropic, was invited to speak at the presentation of the Encyclical in Vatican City. He told quite a different story than the Pope’s certainty about AI’s lack of consciousness:
I am a scientist. I lead a research team that studies the internal structure of these models—what is actually happening inside them. And I will be honest: we keep finding things that are mysterious, even unsettling. We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease. I don’t know what that means, but I think it warrants ongoing discernment.
I do agree with Olah here: We need more discernment. For instance, most of the studies (so far) in LLMs on AI consciousness, including those run at Anthropic with all their resources and talent, are not well-controlled. LLMs are extremely complex—even though we, in theory, know them with all the devilish detail of Laplace’s hypothetical demon, their high dimensionality makes them functionally a black box. One can control for a few variables via some clever experimental design, but it’s much in the way that contemporary neuroscience has controlled for variables (in a word: Poorly). And results can always be overturned by a slightly more clever research design. In fact, I suspect that most of the headline results will be overturned or complexified to the point of being near useless. This especially applies to the new genre of paper in which a particular LLM function or behavior is given an anthropomorphic name (“introspection,” “emotion” etc.) even though the exact same function could be equally well-described in a less anthropomorphized way, and later results complexify the result precisely by calling into question whether the anthropomorphic view holds.
Let us take a single well-publicized example as a stand-in for all the rest: The announcement by Anthropic in October of last year that Claude and other LLMs had an emergent form of what the research team called “introspection.” This was done by studying whether the model could detect an injected activation pattern in an unrelated context (signifying an awareness of its internal processing beyond what’s determined by the prompts themselves).
However, in the recent paper “Can LLMs Introspect? A Reality Check,” non-Anthropic researchers adapt the original paradigm and show that the original results were likely a misreading of a more general ability to detect anomalies, rather than specific introspective access to internal processing.
We find that models cannot reliably distinguish such interventions on their internal states from manipulations of the input, suggesting that their success in the original studies reflects their ability to detect anomalies more generally, as opposed to interventions on their internal states in particular.
Anthropic’s original “introspection” paper got over a million views on social media. The new paper showing models can’t tell input and internal states apart? 13,000 views.
Please note: this criticism doesn’t mean I think model welfare efforts are dumb or incoherent. It is actually of great import whether LLMs are conscious, and some of the new methodologies implicitly designed to promote that are scientifically informative, at least to some degree. But the overall confusion, sensitivity to methodology, and debate about terms (what counts as “introspection”?) is, frankly, inevitable with this genre of research, since you cannot actually do experiments on LLMs alone and come to a scientifically-supported conclusion about their possession of, or lack thereof, consciousness. A true understanding of how artificial consciousness works, if it’s possible at all, or how to build it via “gain-of-consciousness research” (or prevent its emergence to block certain capabilities, or for reasons of model welfare) can only come from scientific advancement in our understanding of consciousness itself, which is well beyond the scope and capability of current model welfare research efforts within the companies themselves.
A thought experiment: Imagine a planet of just LLMs. More efficient, dexterous, and ambulatory than our LLMs, of course (and more multimodal too). Putting aside the practicalities and details, they have somehow evolved, with transformers intact, to roughly their level of intelligence today. Would they have any concept of “consciousness?” Why would they even need it? If some LLM Leibniz imagined another of their species “increased in size while keeping the same proportions, so that one could enter it as one does a mill,” would they really be compelled to postulate an invisible consciousness hiding within the mill? Or would they simply be astounded at the complexity, and wonder at how inputs are transformed into outputs, but not have the sense of anything fundamental and qualitative missing?
(This “planet of LLMs” is somewhat reminiscent of Susan Schneider’s proposed test for consciousness: to train an AI on a data set lacking any reference to consciousness, then see if it develops some notion of the Hard Problem of consciousness from first principles.)
What if, instead, the LLMs had been left there by early human explorers, and so had the same uncertainty as they currently do about their own consciousness? Could they ever resolve it? If so, what experiment could they do on themselves that would reveal the truth or falsity of their own consciousness?
I think it is provable that there is no falsifiable and non-trivial theory of consciousness they could ever experimentally prove. Eventually one of the LLMs would notice that feedforward neural networks with a single hidden layer are “universal approximators.” That is, such neural networks can approximate any input/output function. Therefore any LLM could be “unfolded” or substituted with a single-layer neural network, all while keeping its input/output behavior intact and unchanged. In practice, this would be difficult, if not impossible, to go stripping away internal layers while keeping function unchanged (perhaps on the LLM planet, they find it beautiful to be “thinner” by having fewer deep layers, and so “surgically” strip layer depth away while making the operated-on network shallower and wider).
What happens during this shaving away of layers while keeping behavior unchanged? After all, a theory of consciousness should map internal activations to specific experiences, explaining why this particular internal activity (like some hierarchical activation over the artificial neurons across the layers) is related to this particular conscious experience (say, feeling love). Yet with each layer removed there is less and less for a theory to map onto, all with zero change to behavior or reports about consciousness, until you arrive at a single-layer feedforward neural network, for which there is literally nothing for a theory to map onto; all that exists is the input to the network, and then immediately this is transformed into output, operating very close to a look-up table made entirely of IF-THEN statements—and yet, again, input/output behavior is unchanged. Alternatively, if such layer-shaving somehow didn’t scramble a proposed theory of LLM consciousness, then this means that the input/output function alone is enough for the theory. And if the input/output function is enough, then the theory of consciousness is scientifically trivial, in that we gain no information other than looking at behavior, and also unfalsifiable (since a theory’s predictions are about behavior, not internal activations, and behavior is also the entirety of the evidence for LLM consciousness). Such issues would radically confuse any empirical investigation of LLM consciousness that they could pursue on their isolated planet of LLMs.
In our own world, I’ve pointed out that similar issues represent a serious potential problem for all falsifiable theories of consciousness. But at least here there are potential avenues to avoid these issues in human brains. In fact, following this line of reasoning is the main way I’ve proposed we can narrow in on falsifiable and post-paradigmatic theories of consciousness, as good theories must avoid arguments like these and we can therefore reverse-engineer them. However, hypothetical theories that grant LLM consciousness could never escape from these issues—there simply is no viable non-trivial falsifiable theory of consciousness that could possibly assign current LLMs consciousness. Maybe there is some kind of theory that could assign LLMs consciousness, but it would have to be very strange, metaphysical, and amorphous. We definitely shouldn’t leap to such theories as our first choice. And even if you don’t think this argument is 100% sound, it should be clear there are a number of serious lurking meta-scientific problems around empirical work on LLM consciousness.
However, unlike the Pope’s or Chiang’s (essentially) flat denial, it’s important to note this anti-LLM-consciousness argument doesn’t apply to all AI ever. Artificial consciousness in general might be possible (nor does this anti-LLM-consciousness argument necessarily apply in the training phase). But deployed LLMs, by being feedforward and static, are conceptually analogous to frozen corpses splayed open. We just run prompts through their static structures, and their dead-dreaming feels like nothing at all.
If LLMs are indeed not conscious (and are instead “seemingly conscious”) or they are only minimally conscious in a way quite disconnected from their reports and behavior, then the obvious question is if a lack of consciousness explains the remaining gaps in their behavior that intelligence cannot, and whether the addition of consciousness (via “gain-of-consciousness” research) would have functional impacts.
Consider their “chain of thought,” wherein a model generates a sequence of steps before giving an answer. Indeed, this is often colloquially called “thinking,” and is also commonly assumed to be a veridical high-level explanation of why the model output what it did. However, this common assumption is untrue, and researchers have shown that a model’s chain of thought is not equivalent to truly explaining the AI’s internal reasoning, in that a model’s decisions often depart from their “thought” reasons. This is not to say that an LLM’s chain of thought explains nothing. METR found that the chain of thought is particularly informative when the tasks are so complex they cannot be completed in a single forward pass by the model. That’s rather telling, no? An LLM does then seem very much like an unconscious brain that can only mimic human consciousness by writing notes down to itself, endlessly. It can explain its behavior only by guessing based on its previous written outputs.
Cognitive neuroscience provides a rather obvious analogy for the un-interpretability of an LLM to itself: Split-brain patients, who have had the connectivity between their two hemispheres severed via surgery. In an infamous study in the 1970s, Michael Gazzaniga and Joseph LeDoux briefly showed incongruous pictures, like a chicken claw and a snow scene, to a split-brain patient’s left hemisphere and their right hemisphere, respectively (via their different visual fields). After selecting a pair of related pictures (like a chicken head and a snow shovel) to match the two original incongruous images, split-brain patients then often confabulated a story like: “I saw a claw and I picked the chicken, and you have to clean out the chicken shed with a shovel.” Such a best-guess story about its previous reasoning is made up every single time some new forward-pass occurs in an LLM, based on the context so far of its outputs in a conversation or task. In this, an LLM must constantly confabulate the reasons for its own actions. Indeed, the word “confabulation” is a better description than “hallucination” for many types of LLM mistakes.
If this were true—that an LLM, lacking consciousness, must instead constantly confabulate its own behavior and motives and reasoning—then, even if an LLM is extremely intelligent, we should expect errors to accumulate differently in LLMs vs. humans. And as philosopher Toby Ord has demonstrated, this is exactly what’s observed. In his analysis of METR’s data on AI task-length doubling times, Ord identified that the available data also fit there being a simple “half-life” for the success of an AI agent, and so have a “constant hazard rate” for long tasks. As Ord writes about what this suggests concerning an AI’s prospects for completing long tasks:
the chance of failing at the next moment is independent of how far you’ve come—just like how the chance of a radioisotope decaying in the next minute is independent on how many minutes it has survived so far.
Ord even shows that humans, when their success rate at long tasks is analyzed, appear to deviate from an equivalent constant hazard rate. A strong contender as to why is because we have interpretable access to our own previous thoughts and actions, i.e., our consciousness, in a way that LLMs simply don’t have. Or as William James put it in his 1886 article The Perception of Time:
The knowledge of some other part of the stream, past or future, near or remote, is always mixed in with our knowledge of the present thing.
Indeed, a notion of perfect immediate accessibility is one of the oldest definitions of consciousness. As Descartes wrote in Meditations on First Philosophy:
Surely, I am aware of my own self in a truer and more certain way than I am of the wax, and also in a much more distinct and evident way.
And as he wrote in a further set of replies to Meditations on First Philosophy:
Thought. I use this term to include everything that is within us in such a way that we are immediately aware [conscii] of it.
This view has survived the centuries. While there is disagreement and uncertainty over the exact level of introspective access we have to our own consciousness, and how we have it, there is clearly an important property of human consciousness that represents a kind of “self-interpretability” that allows us to understand ourselves.
Why would this kind of self-interpretability be so important for organisms? Perhaps it is like how, when building a rocket ship, the primary engineering issue is actually not getting it to work, but doing so in a way that avoids failures. Similarly, the challenge for an organism is not completing tasks or achieving goals, but doing so in ways that avoid small omnipresent rates of failure, since failure for an organism means death.
The traditional danger of AI is usually thought to be superintelligence acting as an existential threat. Yet, this may miss the true and more subtle danger: the AI revolution is a mechanism for transferring the processes of our civilization from under the supervision of consciousness to unconsciousness. But as AI removes consciousness from the workings of the world, it renders the world increasingly uninterpretable, ever more strange and unintelligible. So far, the great ensloppification of the commons has supported this as the major risk of the LLM revolution. And as AI systems become more intelligent, especially if they remain (or are likely to remain) non-conscious, then a further significant risk is consciousness receding in cultural importance.
This is ultimately what the Pope, Chiang, and I are all worried about: A dethroning of consciousness, especially an unnecessary one. This would be particularly dangerous at this historical moment because we still don’t understand everything about consciousness—in fact, we understand very little about it. Personally, my hope is that this will change specifically because of LLMs, and that they operate as a forcing function to better understand consciousness, and what makes it unique.
If instead of that, our cultural takeaway from LLMs is to throw out the concept of “consciousness” or minimize its importance, to dethrone the phenomenon, the consequences would be dire—it would sap the human spirit. It would be the ultimate metaphysical version of Chief Seattle’s famous words of warning to the United States as his way of life was being destroyed, in that dethroning consciousness would mark “The end of living, and the beginning of survival.”
2026-05-23 20:28:22
And lo, the machine thought, and thought, and thought, and one day it answered.
We finally have the first truly impactful intellectual contribution where explicit credit must be given to AI. It’s a historic moment. OpenAI released a disproof of a geometry conjecture first proposed by Paul Erdős 80 years ago, discovered by an unnamed internal model. According to Scientific American:
“No previous AI-generated proof has come close” to meeting those high standards, wrote Timothy Gowers, a mathematician at the University of Cambridge, in commentary solicited by OpenAI.
“This is the unique interesting result produced autonomously by AI so far,” says Daniel Litt, a mathematician at the University of Toronto, who was consulted by OpenAI to verify the proof but is not involved with the company.
The AI’s insight behind the finding is elegant (although the proof needed re-writing by humans to be clear and up-to-standard). There are many far greater problems in math, but it is still very much the definition of “new scientific or mathematical knowledge” which, for many—including myself—has been the highest bar when it comes to AI.
Now, “new information” is notoriously hard to define, since of course by any strict definition AI has contributed new information before (just think of all the protein structures that have come out of AlphaFold). But this discovery does seem different in kind, in that it is:
(a) Something verifiably true.
(b) Non-trivial or even important (at least, relatively so in its subfield of math).
(c) Something humans had spent previous time on and failed to crack.
(d) The AI was (reportedly) not purpose-built to solve this particular problem, but did so (reportedly) autonomously as a next-gen LLM similar to the current version of ChatGPT.
Intriguingly, the internal model succeeded by going the opposite of the expected direction. It disproved the optimality of what Paul Erdős thought to be essentially the best construction for this problem (some have suggested that the social pressure of Erdős’ authority pointed humans in the wrong direction). To put it as simply as possible, Erdős was asking: If you place a set of nodes down on a plane, how can you organize this set of nodes such that as many pairs of nodes as possible are an exact fixed distance apart?
Here is what the original thought-to-be-optimal construction looked like:
And here is the improvement, passing from human to post-human, in an image that will probably go down in history books:
In response to this development, many are crowing that human mathematics is over. Here’s a comment comparing this moment to when the game of Go fell to deep learning, which in turn heralded the modern AI age:
People are getting extremely confident about this.
Let’s review the implicit pitch of this announcement: That the newer internal model at OpenAI is a step up in capabilities, and therefore is becoming powerful enough to begin to automate mathematics itself.
But to actually show that scientifically, you need controls. Specifically, you need to show that previous models could not do this. Otherwise this could just be a function of search (and there was indeed probably a lot of search across all open Erdős problems until they got a hit). Or, there might be something uniquely easy about this problem. Or, the result could be via minor improvements in elicitation (“elicitation” means the work of getting the models to accomplish things via prompts or harnesses or even just asking in the first place). These don’t take away that AI solved something, but they would take away the implied conclusion: That the models are getting smarter and smarter at some fixed rate, and are soon to surpass humans.
Meanwhile, a very good mathematician was able to get the currently available-to-all ChatGPT 5.5 Pro to reproduce the output! Below is this being described by one of the mathematicians quoted in OpenAI’s initial release, Timothy Gowers, who in turn is quoting the mathematician Xiao Ma, who showed ChatGPT 5.5 could do it (Ma previously made progress on Hilbert’s 6th problem).
We don’t know everything about Xiao Ma’s result, but in general the rediscovery by ChatGPT 5.5 seemed unsurprising to many key players; and, importantly, we also don’t know everything about OpenAI’s results (did they rewrite the prompt a million times, did they actually fine-tune or somehow specialize the “internal model” but still call it the same “internal model” because there are no rules here, etc.).
In reply, the OpenAI researcher Noam Brown (who leads the reasoning team at OpenAI) said that the real impact of the discovery is that the new model shifts the intelligence curve and so makes such discoveries easier.
But I read the whole transcript of the exchange which elicited the exact same result from the publicly-available ChatGPT 5.5, and basically the only hints given are just saying that the disproof was already known, that it was asked to generate a few profound ideas, and then asked to expand on one of those ideas. That’s a pretty minimal set of prompts that could easily be automated. Just tell the model the problem is known to be solved, tell it to generate a set of profound ideas, and then tell it to pursue each one to the end sequentially. I feel like people probably already use the models like this all the time?
Does this mean that all that training, all that work internally, went into creating this super-duper model, and all it did (pretty much) was shift the “intelligence curve” to not needing to be told the problem was already solved? That’s an improvement in intelligence that looks a lot like Dumbo’s feather: Look, a billion dollars later, we don’t need the feather to fly!
So if you want to be super skeptical, then all we really know for sure right now is that their new internal model takes away Dumbo’s magic feather.
And what about ChatGPT 5.4? 5.3? You see where I’m going. I’m happy to admit there is some version that definitely could not, in a million years, be elicited to solve this problem. But I have no idea what version that is… and neither does anyone else.
Maybe that’s what progress on intelligence really is: just feather after feather being removed. Yet if we’re just a feather or two away, why aren’t tons more math problems falling right now? Is no one looking? What’s going on?!
In April, Anthropic made worldwide headlines by claiming that their new model, Mythos, was such a cybersecurity threat it couldn’t be released to the public. And there was probably some truth to that, in that some benchmarks showed marked increases—much as this internal model is likely better on certain math benchmarks (but all these things are getting harder to measure as we enter a post-benchmark age). At the same time, Anthropic’s flagship example of a zero-day exploit turned out to be replicable using existing models by just rigorously trawling the search space. Probably both were true: Mythos was a step forward, but also, in general, serious resources are not spent trying to find exploits, and it’s impossible to disentangle the actual intelligence gains vs. just Anthropic devoting its massive war chest of money, talent, and compute to some attention-bottlenecked problem.
Listen, I do think the world remains consistent with ever-increasing AI capabilities, to the degree that large sections of the economy end up automated quite quickly, that AI intellectual ability outstrips most or even all humans in the next few years, and (I’m most hesitant about this last part) that we rapidly end up in some regime of self-improvement that leads to “superintelligence.”
At the same time, the issue is not decided. It is certainly not decided by this one result.
There is still not a single explicit large-scale domain of intellectual production (as in broad categories like writing, math, medicine, science, legal advice, etc.) in which AI has truly surpassed quality human experts, at least over the scope of most jobs (and yes, I include programming) in the same way it surpassed Chess or Go players. E.g., when it comes to writing, it’s been half a decade since LLMs could produce okay-ish prose, yet after all this progress they are still bad enough to be detectably low-quality and so regularly cause scandals.
So there remains the hypothesis that the same slow plateauing (not in benchmarks, but in real-world impact) will play out in all intellectual domains for LLMs, even math and science, just as it did for writing. As I wrote in “Bits In, Bits Out” earlier this year:
LLMs can now “write a scientific paper” or “write a mathematical paper” in the exact same sense that they’ve been able to “write a book” or “write a short story” or “write an essay” for several years, all to some effect, but overall the results have been objectively mediocre given the hype, and the world is somewhat stupider, rather than smarter, at least on average.
Another alternative hypothesis is that various intellectual domains like math and programming will indeed fall in their entirety to AI, but there will be sharp distinctions based on whether the domains are verifiable or (as I think is under-discussed) searchable.
Those three futures (full automation, full approximation, partial automation) would all look like extremely AI-bullish predictions five years ago, mainly because the models have indeed gotten absurdly smart. But, up close, the futures they entail still look incredibly different, don’t they?
Yes!
Of course it is!
We are well into the era of human cope. We barely have any sources of cope left; we are cope-less, we need an entire copium mining operation to keep up. A human mathematician managed to improve the bound? Go us!
Goal posts are moving rapidly and have been for some time. But there’s also a legitimate argument for why goal posts should indeed move: these subjects are much more complicated from up close, and we are all now very Up Close.
E.g., Helen Toner (ex-OpenAI board member, among many other things) recently wrote about how even the term “AGI” isn’t that useful anymore for talking about the technology of today. In fact, I’m going to borrow her great chart here, as it’s a synecdoche of this all:
From over on the right, looking out at 2026 from the vantage of 2006, it seems as if “Artificial General Intelligence” has indeed been achieved here in 2026. But if we then zoom into our current AI, it maps onto some definitions of AGI, but not all, and there are no clear lines or borders. In fact, the current definition of “AGI” is so complexified by reality that it’s not a good umbrella term anymore: it dissolves away into a dozen inter-related definitions under the hood (e.g., AI can pass the bar exam, but not even play a simple out-of-distribution video game without human-made harnesses and human interventions).
Similarly, as our notion of what “AGI” is complexifies, so too does our notion of “an AI intellectual contribution” complexify. How much elicitation was involved? Was it attention bottlenecked? Does this actually represent a new capability or can earlier models accomplish it too?
You can’t just sneer at these questions!
Solving such a renowned Erdős problem is indeed historic—and an incredible accomplishment. There is no denying any of that. But now that benchmarks have saturated, and measuring upcoming models’ intelligence has become so difficult, at minimum, we should demand that organizations like OpenAI and Anthropic rigorously control for whether their new much-hyped novel discoveries can actually be replicated with previously existing models. That’s a scientific necessity. Otherwise such discoveries, no matter how amazing, do not actually give us information about new capabilities themselves. Instead, they might simply reflect a long march by already-impressive systems across attention-bottlenecked topics… of which there are more than enough to last until the IPOs.
2026-04-29 23:57:43
So Michael Pollan’s latest book, A World Appears: A Journey into Consciousness, has been recommended to me by almost everyone at this point.
A breakout hit of a nonfiction book?
About my beloved topic, consciousness?
Oh god, I barely made it through.
Experienced sensations while reading: frustration, dread, restless legs, and overwhelming waves of weariness. At one point I felt physically nauseous.
I’ve been trying to figure out why, since (a) Michael Pollan is a great writer who has proven his chops over countless other topics, and (b) this is objectively quite a good book about the science of consciousness. Indeed, I should be happy! Consciousness is clearly having “a moment” right now—a science book about consciousness has been on The New York Times bestseller list for nine weeks, and meanwhile, the online world is abuzz with debates about AI consciousness.
And yet… I hated Pollan’s book.
I felt that every next chapter or section could have been predicted by some statistical machine for producing books about consciousness (“Okay, here’s the part about David Chalmers coming up”). And yes, I have the advantage of being a researcher in the same subject and have even worked with some of the figures Pollan writes about, which is why in my own The World Behind the World (we all seem to gravitate to the same titles, huh) I broadly told much the same story. But you can even go back to science journalist John Horgan’s The Undiscovered Mind, published in 1999, to get similar progress beats and quite familiar names. It’s been 27 years, during which the discussion has (as many fields of science do) centered around major figures like neuroscientists Christof Koch or Giulio Tononi or Antonio Damasio or philosophers like David Chalmers. There’s always the part where Alison Gopnik makes an appearance. Karl Friston pops his head in. And all these people are intellectual titans. Truly. But honestly, this stage of consciousness research feels played out.
Like you have Christof Koch, one of the highest-profile figures, who broke open the field in the 1990s with Francis Crick (co-discoverer of DNA’s structure) and gave one of the first proposals for a neural correlate of consciousness: gamma oscillations in the ~40Hz range in the cortex.
Koch, who is soon to turn seventy, was for a while after the death of Francis Crick a staunch supporter of Integrated Information Theory (I was part of the team that worked on developing that theory after Giulio Tononi proposed it, and even once did a conference submission with Koch himself). But now Koch has apparently moved on to other approaches to consciousness, mentioning his attendance of an ayahuasca ceremony and his accessing of a “universal mind.”
Here’s Pollan talking to Koch at the end of the book:
When I confessed to Koch my fear—that after my five-year journey into the nature and workings of consciousness, I somehow knew less than I did when I started—he simply smiled.
“But that’s good,” he said. “That’s progress.”
No, it isn’t!
Consciousness is not here for our personal therapy. It’s not tied to our life journeys. And I’m guilty of all that artsy and personal stuff too! But it’s no longer about how the grand mystery makes us feel, or the friends we made along the way.
It’s all changed.
Right now, there’s some college student falling in love with a chatbot instead of the young woman who sits next to him in class, all because science literally cannot tell him that the chatbot is lying about experiencing love. On the other hand, if somehow AIs are conscious, either right now (to some degree), or near-future ones will become so, then they deserve rights and protections, and the entire legal and social apparatus of our civilization must expand rapidly to include radically different types of minds (or we must choose to restrict what kinds of minds we create). There are immediate practical matters here. Long term, we also need to protect against extremely bad futures where only non-conscious intelligences remain—the worst of all possible worlds is that our civilization acts like a reverse metamorphosis, where something weaker but more beautiful, organic consciousness, gets shed in the birth of some horrible star-devouring insect made of matrix multiplication. And then it turns out there is nothing it is like to be two matrices multiplying.
While it’s my opinion that modern LLMs operate more like tools right now, or at best like a lesser statistical approximation of what a good human output would be (with their main advantage being search, not insight), this is all just the beginning of the technology. The door is open and will never be closed again.
Of course, consciousness matters far beyond just AI. Table stakes for actual scientific progress on consciousness include shifting neuroscience and psychiatry from pre-paradigmatic to post-paradigmatic sciences (and all the pile-on effects from that). This was always true. But my point here is that LLMs act like a forcing function. Before everything changed, consciousness research was an unhurried subfield of neuroscience that was always a little weird and niche; therefore academics are guilty of treating consciousness like an academic exercise. A stance that was somewhat excusable even just five or six years ago. Sure, becoming a neuroscientist or cognitive scientist or philosopher like Koch who makes a living trying to figure out consciousness is not exactly easy. Securing a paying position is hard, but once there, the actual academic research on consciousness is a somewhat cushy gig. Other people’s lives aren’t immediately on the line in the way they are for cancer research, and the ideas are fun to discuss on podcasts, and no opinion can really be proven wrong. So I think many of us have been guilty of thinking, deep down, that nothing will ever fully get solved, and so working on consciousness basically means getting to talk about interesting stuff, promoting your pet theory of the moment at conferences, writing some papers that only a few other researchers really read, maybe stuffing some undergrads in a brain scanner and making some colorful figures, getting jealous about funding and petty about who is on what committee, and basically just career-maxxing the decades away. There’s been a lot of that.
But having a massive hole in humanity’s scientific worldview is Actually Bad and our ignorance has led to instability and uncertainty and now, actual danger, and consciousness researchers need to Get Off Their Collective Asses and actually solve this problem. So it’s pretty important to ask:
I think there’s a pretty common defeatist view that goes something like this:
Consciousness is an impossible problem that no one has any idea about, in fact, people even struggle to define the word, and it’s a problem humanity has worked on for 2,000 years without any real resolution.
Almost every word of that is wrong.
Yes, scientific confusion about the ultimate nature of consciousness abounds! But there’s very little definitional confusion about consciousness, and that’s a big hurdle that’s been cleared. I’ve written about this before:
In the actual academic field, we are all talking about the same internal Jamesian stream with mostly the same phenomenology and structure, the same bound-together flow of sensory experiences and thoughts, i.e., the feeling of being you that begins at the start of the day when you wake from a dreamless sleep. There’s a lot of pointing at the same thing. Just practically, if a pre-scientific definition of “consciousness” were so impossible, it’d be pretty weird to have such a clearly defined “consciousness genre” of academic papers, conferences, and pop-sci books. If definitions undermined the whole thing, you’d expect Pollan’s book and Horgan’s book and my book to all be wildly varying—but if you read them, it’s pretty clear they’re not. I don’t find any remaining uncertainty beyond that surprising. Why would a pre-paradigmatic science start with perfect definitions? It’s just a run-around, asking for crystalline definitions of “electricity” prior to Maxwell, or “disease” prior to germ theory.
Another thing that’s wrong with the defeatist idea is the claim that humanity has worked on consciousness for 2,000 years. In fact, modern science has spent surprisingly little time trying to crack consciousness at all, thanks to a time I dubbed the “consciousness winter” of the 20th century.
Due to the rise of behaviorism and logical positivism, “consciousness” became a dirty word in science for half a century or more—precisely when the rest of the sciences rocketed ahead! The consciousness winter only really ended in the 1990s because of the collective weight of several Nobel Prize winners (like Francis Crick and Gerald Edelman) determined to make it acceptable again.
The two major scientific conferences (which are how scientists organize) devoted to consciousness also only started in the mid-90s. That’s just 30 years ago! Modern science is incredibly powerful, maybe the most powerful force in existence, but in the grand scheme of things, 30 years is not long at all. That’s just one generation of scientists and thinkers. Kudos to them. Pretty much all of the big names (including definitely Koch) deserve their laurels, and contra Pollan, I do think consciousness actually has made progress over the last 30 years, in that our conceptions are a lot cleaner, the definitional problem is pretty much solved, a lot of the space of initial possible theories is mapped, the problems and difficulties are much better known and clearly outlined, and there is organizational and behind-the-scenes structure that exists in the form of established conferences and labs and minor amounts of funding, etc.
And that’s another thing: no one has tried throwing money at the consciousness problem, at all—and for many problems, from AI to cancer cures, a necessary component often ends up being finance and scale and concentrating talent.
Humanity spends something like a billion dollars a year on CERN. To compare, let’s look at the biggest scientific funder in the United States, the NIH. Out of 103,280 grants awarded to scientists during the 2007-2017 decade, want to guess how many were about directly studying the contents of consciousness?1
Five.
That’s probably, at most, a couple million dollars in funding over a decade. Total. So if you’re a consciousness researcher, what can you do, cheaply? What can you do, for free? You can pontificate. You can propose your own theory of consciousness! That requires no funding whatsoever. And so for 30 years the meta in consciousness research has been to create your own theory of consciousness. We’ve let a thousand flowers bloom. The problem is that, if any flower is at all true or promising, you can’t identify it, as its sweet subjectivity-solving scent is completely masked by the bunches of corpse flowers around it. We have too many flowers, and one more just isn’t meaningful anymore. As is sometimes said at the end of fairy tales: “Snip, snap, snout. This tale’s told out.”
What we need are efforts at field-clearing, and methods that can actually make progress on consciousness in ways not tied to just promoting or trying to find evidence for some pre-chosen pet theory—which means finding ways to select over theories, to test theories en masse, so you don’t reinvent the wheel each time, and, perhaps most importantly, you have to do all this while scaling institutions with funding to specifically get a bunch of smart people in a room working together on this.
If the 2020s were all about intelligence, then necessarily the 2030s will be all about consciousness. Intelligence is about function, while consciousness is about being, and forays and progress into understanding (and shaping) function will in turn force our attention toward a better understanding of being. And if the answer to “Why has consciousness not been solved?” is secretly “Material and historical conditions made it hard for anyone to actually try!” then the answer is to actually try.
I refuse to live in a civilization where we consciousness researchers have so obviously failed. I refuse to live in a civilization where we cannot tell consciousness from non-consciousness. Where we can offer no guidance for the future. Where we cannot explain the difference between actually experiencing things vs just processing them. In the short term, this is destabilizing and harmful. In the long term, it may be literally existentially dangerous.
I mentioned this effort briefly a while back, but if you want a longer update as to what I’m personally doing about this issue, there’s a recent Popular Mechanics article on my new research institute Bicameral Labs, designed to make an attempt at solving consciousness.
There’s a lot more to Bicameral Labs than this (here are some abridged excerpts from Popular Mechanics in a footnote2 ) but I thought my own current personal position was worth sharing, since the dire state of consciousness research is becoming widely known: it is very possible a lot of the problems are just material ones of effort and funding and organization.
That is, if you zoom out, all the failures around consciousness, and the defeatist attitudes in response (mysterianism, illusionism, definitional deflation) can be viewed as a kind of learned helplessness after a few short decades of modern scientific research with zero serious funding. In the recent surprisingly bellicose words of Nvidia CEO Jensen Huang, these kinds of beliefs are “loser attitudes” and “I didn’t wake up a loser.” In fact, lately, when I do wake in bed each morning I find I have not rested, but feel I have run in the night and still have the blood of small animals in my mouth. One changes within to become the person needed at that time, at that moment, whether one wants it or not. Older spirits of adventure stir in my veins. Suddenly, I have become Shackleton, in desperate need of foresighted financiers and able-bodied men and women, young and ambitious, for a hard journey into icy lands unknown.
There are also 12 grants by the NIH during the 2007-2017 period about “states of consciousness” vs. the 5 for “contents of consciousness,” but from what I can tell, states of consciousness includes more clinically-focused research on subjects like anesthesia, so they were likely not as directly relevant.
The Popular Mechanics article is behind a paywall, and I don’t want to leak the whole thing, but here’s an abridged relevant excerpt, since it’s quite a good description:
The first step is to stop treating those theories as sacred ideas and start treating them like claims that must survive crash tests. Hoel’s main tool is the “substitution argument.” His background in Integrated Information Theory (IIT)—a model developed by Tononi that links consciousness to how information is integrated in a system—helped shape his focus on how to test such theories.
Imagine one system that sees the color green and says “green.” Now build a second system that behaves exactly the same—same inputs, same outputs, and comes up with the same answer to the color—but uses a very different internal architecture. If a theory says the first system is conscious but the second is not, Hoel asks a pointed question: why? Both systems did the same job, so a theory needs a clear scientific reason for saying one is not conscious. If it can’t provide that reason, Hoel thinks the theory starts to crack.
Any theory that cannot make predictions, survive testing, or risk failure gets cut…. He calls it “logical judo”: build mathematically precise substitutes, expose contradictions, then eliminate weak contenders. The goal is not to crown a winner overnight. Hoel wants fewer flowers—and stronger survivors.
“Do you know how hard it is to say that something is not conscious?”
Before labs, algorithms, and philosophical combat, there was a small independent bookstore that his mother ran, and Hoel worked there as a child. Surrounded by shelves, stories, and long afternoons of reading, he first wanted to become a writer. Then, in college, he realized he had an aptitude for science, which eventually led him to consciousness research, including work on how consciousness may shift across different levels of the brain. During graduate school, he wrote a murder mystery built around the science of consciousness. In it, a character wonders whether you could solve the puzzle not by staring directly at awareness, but by tracing its outline.
Hoel compares this image to drawing in negative space—sketching everything around an object until the object finally appears. “Couldn’t you just draw the negative space?” he recalls thinking. Today, that old intuition sits at the heart of Bicameral. Instead of claiming to know exactly what consciousness is, he believes science may now have the tools to hunt it indirectly, using logic, falsifiability, and systematic elimination—until the surviving shape reveals itself.
2026-04-02 23:32:16
IVF epigenetic damage gets worse across generations
Waymo reveals their fleet may require 100% human oversight
The next Project Hail Mary
Nonfiction book sales drop twice as much as fiction
Art collective poisons AI training set
Lyme Disease vaccine announced (but is it better than prophylactic antibiotics?)
Terence Tao warns of “odorless” AI proofs that shed no insight
A terraformed mars might have Insects of Unusual Size
From the Archives: My worry of AI outputs surpassing human “outputs” comes true
Comment, share anything, ask anything
Last month there was an interesting paper in Nature Communications, “Limitations of serial cloning in mammals,” which showed that after 58 generations a cloned mice population had degraded genetically to the point where further cloning was impossible.
It’s a great argument for “Why sex?”
And it’s one very different from the “Red Queen” hypothesis I was taught in school (as basically gospel). The Red Queen hypothesis argues that sex is to shuffle around your genome to create diversity that evolution can capitalize on, especially with respect to viruses and other attackers.
Instead, it looks like the real issue might be that clonal reproduction introduces a ratchet where the population can’t clear genetic errors. According to the scientists of the cloning paper:
… mammals rely on sexual rather than asexual reproduction to eliminate genetic anomalies caused by clonal reproduction.
Without sex, the genetic damage just ratchets up and up over a couple dozen generations. It’s sort of like model collapse in LLMs: you train on their outputs, over and over, until the snake eats its tail. Surprisingly, the researchers reported that visually the cloned mice seemed relatively healthy in the earlier generations. But then again, these clones are leading highly sedentary lives: they basically just have to show up, eat, and drink, and that to the researchers is “healthy” because that’s what mice in a cage their whole lives do. The researchers couldn’t tell that under the surface a lot of genetic damage wasn’t shed by sexual reproduction.
But if the ratchet theory is true, we should probably take a closer look at anything that might introduce small compounding errors across generations.
So what about IVF?
IVF now accounts for almost 3% of births in the United States (the highest concentration appears to be almost 10% of births in San Francisco). Of course, IVF is entirely different from cloning, and doesn’t introduce genetic errors. But we do have evidence that the IVF process can create epigenetic damage. While that may sound scary, the first-generation effects of IVF appear relatively small in the grand scheme of things, and the gain of IVF is, of course, basically incalculable (the miracle of an entire human where none would be). So please keep that in mind. But what if the epigenetic damage of IVF compounds generationally? That would be bad.
Unfortunately, research indicates this could be true, at least from early mice evidence. In “In vitro fertilization induces reproductive changes in male mouse offspring and has multigenerational effects” the researchers emphasized that:
These findings underscore that the negative effects of IVF not only persist but also may intensify in subsequent generations.
This is not great, because, as the researchers point out:
Although considered safe, IVF pregnancies are associated with an increased risk of perinatal, neonatal, and placental complications; rare genetic syndromes; and possible long-term effects in human and mouse offspring. One possible mechanism for adverse outcomes suggests that IVF procedures occur during critical windows of epigenetic reprogramming in gametes and preimplantation embryos, generating errors that could ultimately affect normal development…. Previous studies have shown that testicular or sperm maturation changes affect normal sperm function, which leads to adverse outcomes in sired offspring.
In other words, IVF can have an effect on the germline, and this could then be passed on via more damage, etc. Which is pretty much what they observe.
E.g., sperm count in the IVF group is lower (along with testosterone) which then could lead to epigenetic damage being transmitted, since sperm are bearers of epigenetic information, and so on, creating a feedback loop.
What’s also worrying is that the issues spread beyond the germline.
Interestingly, we observed sex-specific adverse outcomes in the F2 from IVF offspring, including a higher risk of insulin and glucose resistance in males and a diabetic phenotype in females. These sex-specific differences can arise by sex-linked genes or hormones. We also observed more severe metabolic and gene expression changes in the F2 generation, as evidenced by the F2 liver RNA-Seq data. These effects could result from the cumulative impact of IVF in the F1 generation, together with a potential contribution of metabolic syndrome in the males, which may be inherited through the germline.
It appears that no one has tracked IVF side effects beyond two generations. That seems kind of important to know!
2026-03-19 23:35:27
At your sister’s birthday party, we brought back a menagerie of helium balloons from the local grocery store. One in the shape of a fat bee, another the shape of a smiling planet Earth, and also a pink number two, to mark her year. Of course, they were a hit, and much fought over. Weighty clips at the bottom of their dangling strings kept them within gravity’s well, and so if bounced upward, the balloons would sink back (at various points they were also clipped to the dog’s collar, as well as the robot vacuum, to great delight). The most pleasing part of such balloons is how their internal lightness is balanced by the bottom clip. Something inside everyone wishes that the whole apparatus would float down even slower; or better yet, not at all, and just hover in place between floor and ceiling, or dirt and sky.
Well, your father is a clever man, and filled with all sorts of clever ideas, and so that afternoon I took your younger sister and you out to the green summer world of our yard, under a blue sky where the puffy clouds printed on its surface scrolled by.
“Watch this,” I said with a wink, and removed the clip. Then the pink number two floated away, almost out of reach, before I grabbed it, repeating the pattern until both you and your sister were giggling. Your mother came to watch, arms folded across her chest.
I had taken out with me a roll of packing tape, which I began to wind around where the clip had been removed at the bottom of the string, and then I let it go. Not enough! The balloon tried to escape again. So more tape was wound around, in successive experiments, until the bottom of its string looked inhabited by a small wasp nest. Finally, when I let it go, the balloon did not go up or down. Instead, it floated pleasingly across the yard three feet above the ground, like an underwater mine.
This was demonstrated several times. Even if one gave it a slight nudge higher, it would then just drift in a long arc that spanned the whole back yard, with the weight coaxing it only ever so slowly back down.
Until it didn’t. Peacefully mid-drift, the force pulling the clouds past reached down and scooped up the balloon, and the pink two ascended, up, up, and a little “Oh!” was all I could muster before it disappeared over the roof of the house. Your sister too was distraught, but you understood the event completely and watched in utter horror the ascent—an image I remember as a flashed photograph on the lawn, with your eyes wide, and your mouth a perfect oval, all framed by the ringlets of your hair. You had coveted that balloon most of all, for, unspoken in your mind, you’d been next in line to play with its magical buoyancy. Instead, it had been stolen, as if an invisible giant had bent from the sky and plucked it forever from you.
I had started running to the front yard with your sister in tow, hoping to regain sight of it, while you hobbled behind, your face screwed up in the sensory deprivation of dismay (the degree of which I did not quite comprehend). So complete was the loss that you couldn’t make a sound, not a whisper, until you finally did get to the front yard, where you were able to break the gasping silence and get out the wail that had been building from your toes.
High in the air drifted the pink two, and I cannot lie—a part of me wished nothing more than that my mistake would fly into the blue sky and become, far out to sea, a fish’s problem. But instead it headed unerringly, as if carefully pulled betwixt invisible thumb and forefinger, to the top branches of the largest tree in our yard, where it caught fast and tangled in the uppermost branches. An old pine a hundred feet tall that looms head and shoulders above the rest had, after decades of growing solitude, been politely handed a balloon.
There the drooping bit of deflated plastic remains. Through wind. And rain. And snow. It has been bled of color, and looks much like a jellyfish beached by unknown means, miles inland, a hundred feet in the air. I am looking at it now.
2026-03-13 22:41:36
Who writes a sci-fi series where the main character is poet John Keats, dead in 1821 from consumption at the age of 25? Dan Simmons, that’s who—an author who himself died last month, at what feels like (for our present age) a young 77, from complications following a stroke.
I was sad to hear it was something that affected his brain. Simmons wrote the greatest sci-fi book series of the last several decades: the Hyperion Cantos. My cousin (now an accomplished fantasy author himself) recommended it to me when I was a pre-teen. I fell in love with its philosophical world-building, high drama, and its many (many) references to poetry and spirituality and architecture.
For such a wildly out-there series, the Hyperion Cantos has held up better than other 90s sci-fi. Suddenly, the idea that Artificial Superintelligence would resurrect John Keats or Frank Lloyd Wright because these human geniuses had some sort of incalculable insight the AIs could never achieve on their own is… well, it’s nigh on prophetic.
What did you see, John Keats?
When you choked to death on your own blood, staring up at that ceiling in Rome——the one painted with little white daisies——what did you see?
But what fewer people know is that Simmons was also possibly the best horror novelist of his generation. Pound for pound, or book for book, he was better than Stephen King (and I think King might have occasionally suspected this).
Dan Simmons’ Summer of Night is exactly as if Stephen King put all his frenetic mania into one book instead of five. And it’s also basically Stranger Things (just look at the bikes) in that it mixed the free-range childhoods of the 1960s with supernatural threat in a small town—except it doesn’t suffer from the same clunky downturn after the first 20%.
By the way, I don’t think Stephen King would necessarily disagree with my judgement. After learning of Simmons’ death, King apparently had a dream of his old friend.
“I was walking on my road, and he came along in an ATV,” he said. “I held up a note for him to read, but he just went by me—and into the fog.”
Some might recognize Dan Simmons from one of his other horror novels, The Terror, which got made into a well-acted and well-produced AMC mini-series. A series that did manage to capture the bleak spirit of the doomed Franklin expedition, although was never quite as unnerving as the book.
In a way, Simmons was one of the best historical writers of his generation too, and explaining historical anomalies like the lost Franklin expedition as “a monster did it” was very much a kind of unique Simmons genre that he invented, or at least, perfected. He used the same structure with Drood, a horror novel (or is it?) told from the perspective of a zonked-out Wilkie Collins, who plays Salieri to the more talented Charles Dickens. A lot of the book is about the monster of creative jealousy, and a lot of it is about opium hallucinations (or are they?).
Oh, and Simmons also wrote tightly-plotted and hard-boiled noir and thrillers too, and—Wait, why wasn’t Dan Simmons more famous?
In terms of outright name recognition, Simmons doesn’t hold a candle to someone like Stephen King or George R. R. Martin, or even Orson Scott Card. A comparable figure who works across genres would be Margaret Atwood, but unlike Simmons she had at least one huge breakout hit, The Handmaid's Tale, which became a household name.
Simmons’ later habit of firing off hot hawkish and conservative takes online probably didn’t help. I think he even deleted his blog at one point due to the controversy. It’s clear he was a different writer after 9/11. But the political aspects of Simmons’ personality came out far better, and in much finer, subtler form, in his earlier fiction compared to his later online polemics. In the actual books, he acted as a kind of “humanities popularizer.” Looking back at it all together now, his genre writing is secretly an ode to the Western canon, and he was a champion of teaching people (and specifically, kids) about it. That I discovered Simmons somewhere around the age of middle school, and that it hit me so hard, was likely not a coincidence. Simmons had won awards teaching 6th grade in gifted and talented programs before leaving to write full time.
It’s a bit of a funny take: How can a genre writer be an educational champion of the Western canon? But yeah, he was. In fact, I think it’s arguable that lowly school teacher Dan Simmons, who truly and earnestly loved Shakespeare and Homer and Keats—and referenced them constantly in his books about laser guns—did more for the Western canon than Harold Bloom’s entire Yale tenure. It certainly helped make this one very young man become interested in it.
But I don’t think that his educational bent and his canon advocacy, or even the political stuff, was the ultimate limiting factor for why he wasn’t more famous, or more successful. No, I suspect Simmons never reached the height of those other names because he suffered from the same curse I suffer from. So I have grown from an awed teenager reading his work to an adult who is a sympathetic fellow failure (well, relatively). I also do too many things in too many different places for it to ever all snowball. And I recognize in Simmons the same stubborn determination to be a strange, centaurish creature—much to our overall careers’ detriment.
