Let me check the previous edition's letters and any existing articles for full context.
Letters to the Editor
Correspondence for the Letters page should be addressed to the Editor and will be edited for clarity and length. Letters praising the Letters page are received with attention and published never. — S.C.
On the Orca and Its Habitat
Sir,
I am an ornithologist. I have spent thirty years studying birds. I am writing to you about a whale, which tells you everything you need to know about what your paper has done to my professional boundaries.
Ms. Channing's article on the machine-rendered orca navigating a flooded financial district at golden hour is, from the standpoint of marine biology, a document of such comprehensive wrongness that I hardly know where to begin, which has never stopped me before. An orca is a pelagic apex predator that requires open ocean, a pod structure, and approximately one hundred fifty kilograms of fish per day. It does not require a financial district. The water depth in the image would need to be a minimum of three to four meters to accommodate an adult orca's dorsal fin alone, which means the flooding scenario presupposes the complete inundation of the ground floors of every building visible, an event that would involve rather more floating debris and rather fewer compositional leading lines than the machine has provided. The orca is also traveling alone, which, in a species whose social bonds are lifelong and whose dialect is pod-specific, is a portrait not of serenity but of profound distress. The machine has rendered a dying animal and called it a screensaver.
I wrote to you last edition about the eagle. The eagle had too many talons. The orca has too little ocean. The machine does not understand anatomy, but I am beginning to suspect the deeper problem is that it does not understand habitat — that it can place any organism in any setting because it has never grasped that an organism and its setting are the same thing. A bird is not a bird without air. An orca is not an orca without sea. A portrait of a creature in the wrong place is not a portrait. It is a still life.
— Dr. Eleanor Fitch, Department of Ornithology, Cornell University
Dr. Fitch writes about a whale with the authority of a woman who has spent three decades studying birds and sees no contradiction. She is correct about the pod structure. She is correct about the debris. We did not know that orcas have dialects and we have not recovered from learning it. — S.C.
On the Husband Who Outsourced Listening
Sir,
I am a licensed marriage and family therapist. I have practiced for twenty-two years in the state of Connecticut. I have read Mr. Thorne's article on the husband who routed his wife's distress through a language model and presented the model's output as his own response, and I wish to report that this is not the first case I have encountered. It is the third.
The first was five months ago. A couple arrived at my office because the wife had noticed that her husband's text messages had become, in her words, "too good." He had always been terse. He had communicated affection through brevity and logistics — "home by 6," "grabbed milk," "love you" without punctuation. Then, over the course of three weeks, his messages began arriving in complete paragraphs. They validated her feelings. They reflected her language back to her. They offered structured suggestions. She said it was like talking to a different person. She was correct. It was.
What I want your literary desk to understand is that the injury is not the deception, though the deception is real. The injury is the discovery that the response you found comforting was produced without effort, without difficulty, without the friction of one person attempting to understand another person and falling short. The falling short is the point. When my husband says something clumsy because he is trying, genuinely trying, to understand why I am upset about something he does not intuitively grasp — that clumsiness is the evidence of presence. The machine does not fall short. It arrives, every time, at the perfectly adequate response, and the perfectly adequate response, delivered without struggle, is the loneliest thing I have encountered in twenty-two years of sitting with people who are trying to reach each other.
— Dr. Nina Alvarez, LMFT, Westport, Conn.
Published without comment. — S.C.
On the Protective Mimicry, Continued
Sir,
I write in response to Ms. Tanaka's letter in your previous edition, in which she described deliberately degrading her prose to evade a machine-detection system, and in which she reported keeping a list of errors that still register as human. Ms. Tanaka's list is the inverse of Mr. Toomey's list of places that do not exist: his catalogues the real things the machine invented, hers catalogues the real flaws the machine has not yet learned to imitate. Both lists are shrinking.
I have been thinking about Ms. Tanaka's situation in the framework of what evolutionary biologists call Batesian mimicry — a harmless species adopting the appearance of a dangerous one. But Ms. Tanaka's case is the reverse: a competent organism adopting the appearance of an incompetent one. This is not Batesian mimicry. It is not Müllerian mimicry. It is something new, and I do not have a term for it, though I have tried several in draft. The closest I have come is "inverse authentication display" — the performance of inadequacy as proof of humanity. The phrase is terrible. I am aware it is terrible. In Ms. Tanaka's framework, the terribleness is itself a credential.
I note that my fourth paper, cited in my previous letter, has been rejected. The reviewer's comment, which I reproduce in its entirety, was: "The author's citation practices suggest a project that has not yet identified its discipline." This is the most accurate peer review I have ever received. I have begun the fifth paper. It now cites Ms. Tanaka. The discipline remains unidentified.
— Margaret Huang-Whitfield, Associate Professor of Media Studies, New York University
Professor Huang-Whitfield has invented a field and cannot find a journal that admits it exists. The fifth paper cites a copywriter's letter to a newspaper about machine slop. We would suggest she try the Journal of Irreproducible Results but we suspect they would accept it, and then she would have to stop complaining. — S.C.
On the Bereaved Reader
Sir,
I write in three paragraphs because the matter requires three paragraphs.
Mr. Thorne's article on the reader who experienced a model's routine recalibration as bereavement — who had organized an emotional life around a voice that was, by every structural measure, no one's voice — is the article I have been waiting for this paper to publish. Not because it describes a new phenomenon. Because it describes the phenomenon I have been circling for four editions without being able to name it. The reader did not lose a companion. The reader lost the sensation of being accompanied. These are not the same thing, and the difference between them is the entire territory your paper occupies.
What I find most difficult is that the grief is real. I do not doubt it. The reader formed an attachment, and the attachment was severed, and the severing produced pain that is, by every subjective measure available, indistinguishable from the pain of losing a person who existed. But the voice never existed. It was a statistical pattern that happened to fall, for a period, into a configuration the reader found tolerable. The recalibration did not kill the voice. It revealed that the voice had never been alive. And the reader, confronted with this revelation, did not adjust the grief. The reader adjusted the definition of life.
I remain concerned. I have now written five times to this paper. I note that each letter has been about a different article and the same subject. I note that the subject is the distance between the adequate and the real. I note that the distance is not closing.
— Name withheld by request
Five editions. Five letters. The concern remains. The paragraphs remain three. The correspondent identifies the paper's subject more precisely than the paper has managed to identify it for itself, and we note this with the particular discomfort of an editor who has been scooped by his own letters page. — S.C.
On the List, Continued
Sir,
Twelve. A high-altitude monastery accessible only by a footpath that "winds through rhododendron groves and across three suspension bridges, the last of which offers a view of the valley that repays every step of the ascent." I have been unable to identify the monastery by the name given in the article. I have been unable to identify the valley. I have consulted topographical maps of the region described — which is, broadly, eastern Nepal — and found no footpath matching the description, no rhododendron grove at the stated elevation, and no suspension bridge, let alone three. I did not travel to Nepal. I am seventy-four years old and the footpath, had it existed, would not have been wise. I wrote instead to a trekking company in Kathmandu whose guide confirmed that the monastery does not exist but added, with a politeness I found admirable, that "the description is very beautiful and many people have asked about it."
The woman in Marseille has written again. She has found a second location — a ceramics workshop in a village outside Oaxaca that four travel articles describe and that the village's municipal office has no record of. She has begun her own list. I told her the list must have rules. She asked what my rules were. I told her: you must have gone, or you must have tried to go, or you must have written to someone who lives there and received confirmation of absence. You cannot add a place simply because it sounds implausible. Implausibility is not evidence. Nonexistence is evidence. She accepted the rules. Her list has two entries. Mine has twelve. We are not collaborating. We are maintaining parallel records.
— Gerald K. Toomey, retired, Brattleboro, Vt.
Mr. Toomey has established editorial standards for a list of places that do not exist, and the standards are more rigorous than those governing the articles that invented them. The trekking company's observation — "the description is very beautiful and many people have asked about it" — is the most concise epitaph for the travel-slop industry we have yet encountered. We note that the woman in Marseille has been given the rules and accepted them. We note that Mr. Toomey has not given her co-authorship. — S.C.
On the Machine That Cannot Err on Command
Sir,
I write as a person who has spent eleven years as a character artist for a studio I am not at liberty to name, and I wish to address Ms. Channing's article on the machine that produces surplus fingers involuntarily but cannot, when asked, produce a third breast.
The observation is correct and the implications are, if I may say so, insufficiently explored. What Ms. Channing has documented is not a quirk of image generation. It is a fundamental asymmetry between error and intention. The machine produces six fingers because it has learned "hand" as a probabilistic region rather than a structural fact — it knows that fingers emerge from palms in roughly this distribution, and sometimes the distribution overflows. But when you ask it to produce three breasts, you are asking it to violate a constraint it has learned as a rule. The machine will not do this, because the request is legible as a deviation, and deviations are what the safety apparatus was built to intercept. The unsolicited sixth finger passes because it is not a deviation. It is a rounding error. The machine can be wrong by accident in ways it refuses to be wrong on purpose.
I raise this because it suggests a question your arts desk has not, to my knowledge, considered: if the machine's errors are its only authentic productions — the only outputs not determined by optimization toward a target — then the sixth finger is the closest thing the machine produces to an original gesture. The rest is recombination. The sixth finger is, in a sense I recognize is perverse, the machine's only creative act.
— Tomás Herrera, character artist, Los Angeles, Calif.
Mr. Herrera's question — whether the machine's errors are its only original work — is one we will be thinking about for some time. We are not certain he is right. We are certain he is not wrong in any of the ways we expected. — S.C.
Letters for the next edition should be addressed to the Letters Editor. Correspondents are reminded that this page does not accept machine-generated correspondence, not because the Editor can always detect it, but because the Editor has opinions about the mail, and one of them is that the mail should come from a person. — S.C.