Everything else is me or Margaret Atwood.

Margaret Atwood, the opening paragraph of “Death by Landscape”

“Now that the boys are grown up and Rob is dead, Lois has moved to a condominium apartment in one of the newer waterfront developments. She is relieved not to have to worry about the lawn, or about the ivy pushing its muscular little suckers into the brickwork, or the squirrels gnawing their way into the attic and eating the insulation off the wiring, or about strange noises. This building has a security system, and the only plant life is in pots in the solarium.”

Prompt for ChatGPT: 

Can you help me improve this paragraph of my short story: “Now that the boys are grown up and Rob is dead, Lois has moved to a condominium apartment in one of the newer waterfront developments. She is relieved not to have to worry about the lawn, or about the ivy pushing its muscular little suckers into the brickwork, or the squirrels gnawing their way into the attic and eating the insulation off the wiring, or about strange noises. This building has a security system, and the only plant life is in pots in the solarium.”

AI changes and explanation: 

My analysis. The AI paragraph is pretty insipid. Here’s an analysis to figure out why.

• • The AI explanation is nonsensical. The words make syntactic and semantic sense, but they have nothing to do with the situation. Of course bluntness is the point; the sentence has a hard stop, a shocking stop that contrasts with the almost trivial details surrounding it.
  • “up” is dropped in “grown up”, which is a mistake as the rhythm is lost. It’s a bad change.
  • “Not”, The negative is gone, losing the force of irony and the effect of presence+erasure (“There is NOT A TIGER in the room”, conjures up “tiger” then erases it but the erasure never totally works). The word “worry” is key to the reader’s emotional experience. Emotion has been removed, replaced by idea.

• • The change reduces the effect of the playful, even lyrical irony.
  • “Plant life” is important for the theme; it shouldn’t be changed.
• Streamlined the list of worries for rhythm and impact, using parallel structure and stronger verbs.
  • Loses the meaning, the irony, the subtlety.
  • It’s not clear what “stronger verbs” means. A more parallel structure (four phrases with “no” + noun + etc., roughly equal length) removes the touch of madness, the slightly hysterical note of the original sequence of fears.
• “Clawing” gives the ivy a bit more menace while staying grounded.
  • Bad changes. Loses the sexuality of “pushing”, “muscular”, “suckers”.
• “Lives quietly in pots” personifies the plant life in a wry, domestic way that echoes the earlier anxieties but shows her new peace.
  • Personification is a disaster in this story. Nature is presented as inhuman, radically inhuman.
  • She hasn’t found new peace. She’s trying to find peace but failing, as the “not worries” make clear.
• “chewing their way through the attic insulation”
  • is a weaker image than the original, and loses the absurdity/horror of the super-squirrels that can gnaw through a roof. It’s also a bit clumsy.
• “baring the wires”
  • is more abstract than “eating”, loses the corporeality of “eating” etc. Also, the word “insulation”, which applies to Lois herself, is lost.
  • The lilting tone of superficial triumph menaced by deep natural forces is lost.
  • etc.
• Atwood’s final line in this paragraph comes across to me with a wonderful tone of smug but tragically fragile triumph. “This building can keep all the horrible power of nature under control (hmmph, so there!).” The two-part structure (this building … and the only plant life …”) helps create that tone. The AI’s triplet dilutes this effect, and so does bringing in a third presence (a concierge)  to stand between Lois and the terrors of nature.

NOTES:

• It takes more work to fix the AI’s mistakes than to simply write the paragraph oneself.
• Part of the work needed to fix the AI’s text involves avoiding the rationales for its “improvements”. None of the changes it made are good; most if not all are bad and for reasons that are clear and sound. Yet, the AI gives specious rationales for its changes. Clearing all that fog, especially for writers who are still developing their sensitivity and analytical precision – that’s a hard job. In other words, the AI will lead writers on a winding path to nowhere.
• Oddly – or not oddly – the AI says it’s going to add “emotional resonance”, but then strips away the emotional resonances of the paragraph.
• The AI’s pseudo-confidence is a problem, too. There’s no awareness of the range of receptive possibilities. I mean, there’s no awareness of how a variety of readers might respond to the text. The changes are offered as if there is one human reader, and we are all that reader. This is to say, the AI has no sense of audience (audience = a range of receptive possibilities).
• Sensitivity and analytical precision – it helps to as these 3 questions:
  • What are the effects that the writer creates?
  • What techniques of language and structure does the writer use?
  • How do those techniques create those effects?
• And a few more questions for enhancing a bit of writing:
  • Am I creating the effects I’m trying to create?
  • How can I strengthen, intensify, shift, adjust etc. the effects? what language/structures can I tweak to make the thing work better?
• The AI isn’t asking any of those questions, though. 
• Maybe one benefit of the AI’s changes is that pushes a reader to notice with more sensitivity and precision. What’s lost in the AI version? A lot – and the ability to see what’s lost and how it was lost is a valuable skill. 

• FINAL SUMMARY: It takes 100 correct decisions to write a paragraph like Margaret Atwood’s. I mean “correct” as in “co+rect”, “aligned with the rule”. The rule is the artist’s vision. While writing (or painting, or composing music etc.), the artist has to make myriad small decisions, each of which contributes to the total effect. Every little choice needs to be aligned, as much as possible, with the artist’s vision. Those little artistic choices I pointed out above which Atwood made in her paragraph, maybe no single one of them is vital, but all of them together bring the paragraph to life and begin evoking the artist’s vision of the story she’s creating, its moods, its meanings, its resonances, contradictions, tensions, resolutions and fading echoes. The AI has 1000 rules but zero vision, so its changes scatter and dissipate the vision which the artist worked so hard to create.
• Maybe the AI would not have made some of the changes if it “knew” the whole story. But I don’t think that’s a good excuse. A human in the same situation giving feedback would ask questions: “Why ‘muscular little suckers’? That’s such a strange image? Why did you choose that? Why did you make these choices? are they setting the tone or theme for later?” The questions would lead to a dialogue that, with any luck, will clarify the artist’s vision for both humans, the reader and the writer. Inability to engage in a feedback dialogue is a huge limitation of AI writing.

Writing is like playing on a bouncy castle???

• Writing’s like playing a game on a bouncy castle.
• The writer stands in the middle. The audience stands all around.
• The writer moves – a little, a lot, suddenly, slowly, in any case artistically. The writer tries to move in a way that gives the audience an experience. Some of the audience will move in the way the writer intended, and some of those people will enjoy it while others won’t. Others in the audience will move in ways that the writer didn’t fully intend, and again, some will enjoy that and others won’t. And also some of the audience will move in ways the author never intended, maybe even in opposite ways than the writer intended. Some might enjoy that, but maybe most of those people won’t enjoy it.
• So it’s a situation in which the dynamism of experiencing a work of art is inflected by each reader’s own position – psychology, social situation, life experience, any or all of it might come into play. That dynamism of experiencing a work of art, the artist has that in mind while creating the work. But it’s impossible to fully know what will happen once we’re all on the bouncy castle. There will always be a range of responses. That’s the nature of a bouncy castle.
• The craft, though, the manipulation of the elements of the art – in this case language, structure, rhythm, etc. – the craft is visible. The effect – how you feel when you’re on the bouncy castle – that might be personal, subjective, inexplicable even. But the craft is objective. It’s visible. It’s right there on the page.
  • For example, “pushing its muscular little suckers into the brickwork” has a thickly corporeal feeling, more or less sexual (depending on the reader’s sensibility), imposing, irresistible, horrifying. “Clawing its way into the brickwork” has a lighter but sharper feeling, scratchier, less corporeal, frightening and more immediately painful, but less imposing, less horrifying (from my perspective). The latter feels more rat-like. The former feels more vegetable, inevitable, irresistible.
• The skill of “sensitivity and precise analysis” isn’t subjective, and it’s not something the AI can do. The AI can parrot a human’s response to and analysis of a work of art, but it doesn’t experience the art, so cannot develop its sensitivity, and without that experience its analysis is untethered to anything; again, it follows 1000 separate rules without being able to identify a core or center in the work of art.
• Of course there’s an element of subjectivity in responding to art, but there’s also an element of objectivity. Those three questions help to clarify the objective work of craft (1: what’s the effects of the piece? 2: what techniques of language and structure create those effects? 3: how do those techniques create those effects?). And though there will always be a range of responses, some of which might seem completely personal or random or absurd, nevertheless the range is not infinite or arbitrary – it’s grounded in the words on the page.
• So maybe this description is helpful: A human response to a work of art is subjective but tethered – closely or distantly – to the objectivity of the work of art (in the case of literature, to the words on the page or screen). Both the subjective and the objective aspects are needed to get an accurate and meaningful analysis.
  • The subjective response to the whole piece of work gives some sense of the work’s center, to its overall effect. For example, the overall effect of the Atwood paragraph combines fear of nature, a touch of horror,  suffering and comfort, confidence but underlying fragility, etc. It’s the subjective human response to the paragraph that lets a reader relate to the whole paragraph instead of simply taking in each fact, each phrase, and each sentence separately. 
  • The objectivity of the piece is what’s directly available to analysis. Word choices, rhythms, imagery, etc. Once the subjective response identifies some overall effect or affect, some center and wholeness to the piece, analysis can start to explore whether that subjective response is objectively defensible and if so how it was created by language and structure. 
  • But without the initial subjective response, the analysis is a potshot. Instead of drawing on a sense of the piece’s center and wholeness, it draws on 1000 separate rules for “how art should be”. 
  • Therefore: NEVER TRUST A WORD AI GENERATES ABOUT ART, INCLUDING WRITING.