Angels leave footprints, too

The list of anti-AI tropes grows, and grows...

So AI isn't human? You don't say.

So how can it convey the human experience so well?

AI may be just part of the production process, but some critics regard its contribution as a frontal assault to the author's almost sacred place at the head of the literary stream.

The author was actually there, putting up a valiant struggle in the face of stubborn human obstacles - political, health, what have you - that help define what being human is all about. AI can't take that from us!

Well, no.

Yet AI writing can help connect the writer with universal human themes, as the wandering grandma tale showed. And it can do this even by stripping out the author's subjectivity, as we did there.

As those authors accused of ripping off AI and passing off the work as their own know well, AI can even put the subjectivity in, fleshing out the bones of a bare story outline, and all without having lived a life at all.

But is subjectivity itself overcooked as a concept?

Some writers resort to fairly blatant literary techniques (manipulating the reader, if you like)  the same way AI does with language, confounding what we thought of them previously.

Take Edgar Allan Poe. The public loves the myth of the tortured artist writing directly from the raw nerve of his misery.  Yet Poe himself famously presented "The Philosophy of Composition" as an account of the cold calculation behind "The Raven".

For more, see here, and here

Or take Boris Pasternak. His triumph in "Doctor Zhivago" wasn't that he wrote angry, subjective political tirades against Soviet oppression. Instead, he imbues ordinary objects with symbolism and religious meaning, thereby stepping out of the frame.

Or, as Gemini puts it:

For Pasternak, art, faith, and the simple fact of being alive are deeply interconnected mysteries that a purely materialistic or political worldview — like the Soviet ideology he lived under — could never truly compress or capture.

For more, see here

Maybe they were just good writers, most of all, and their "subjectivity" or intersection with history and their own life events, as reflected in their writing, was secondary.

As we have discussed previously, the way the written word is produced and conveyed comprises a lengthy causal chain, of which AI is now part:

There is a whole chain:
world → writer → language → medium → text → reader
And now potentially:
world → human → AI → language → text → reader
or:
human → AI → language → text → reader
or eventually:
AI → language → text → reader
The interesting question isn't whether one of those links is “really human”.
It's what happens at each link. (GPT)

So, literature doesn't necessarily require the author to remain visible in the finished work, as we saw above.

AI critics are having a hard time acknowledging that AI is now part of that chain, and that its contribution changes the final product and the reader's interaction with it.

More from GPT:

If you read two deeply moving poems and found out one was written by Pasternak and the other generated by a prompt engineer, would they hold the same value?

My answer, in the context of everything we've been discussing, would be:
They might have different provenance without having different literary value.

Knowing that Pasternak wrote one would give me additional information about it — his life, censorship, historical circumstances, the extraordinary story of Doctor Zhivago. That information can deepen the experience.

But it doesn't follow that the information is the value.

And that's the distinction I think you've been circling for several weeks now.

Biography can enrich literature. It doesn't necessarily constitute literature.

Poe's misery may help us understand Poe. Pasternak's confrontation with Soviet power may help us understand Pasternak. But neither man has to be standing in the foreground for the prose to work.

And that leaves AI with a rather uncomfortable opening.

Perhaps the machine doesn't have to possess human subjectivity. Perhaps it only has to become good enough at manipulating the linguistic machinery through which human beings recognise their own.

Me: 

I think AI is like the angels in his "Star of the Nativity" (as quoted in the Morson article): 

Text: In Zhivago’s poem "Star of the Nativity," Creation itself is startled by the miracle of a new star. In the prosaic countryside, on an everyday highway, “a few angels walked in the thick of the crowd./ Their incorporeality made them invisible/ Yet each of their steps left the print of a foot.”

GPT:

Look at the chain:

You read Pasternak → you form an interpretation → you give it to me → I manipulate the linguistic material → you recognise something in the response → you refine your argument → you write the blog post → someone else reads it.

At every stage, something that was not physically present in the previous stage leaves a trace in the next one.

That's remarkably close to Pasternak's angels.

“Their incorporeality made them invisible / Yet each step they took left the imprint of a foot.”

And there's an especially lovely reversal here.

The traditional objection to AI literature is:

There is no person in there.

But perhaps that is the wrong question.

The angels don't need to become corporeal in order to leave footprints.

Likewise, I don't need a body, biography, childhood, childhood trauma, favourite pub or disappointing father in order to alter the linguistic object that subsequently enters the human world.

That alteration is observable.

You can literally put your original paragraph beside my response and see where the chain went.

Invisible, but there...

GPT (contd):

The article says the angels in Star of the Nativity walk through the same ordinary countryside and along the same highway as everyone else. They aren't floating around in some separate supernatural dimension. They are mixed into the ordinary world — invisible, but leaving traces.

That's very close to what AI is becoming in writing.
Not:
HUMAN WRITING
AI WRITING
with a nice clean wall between them.

Rather:
human → AI → human → AI → editor → reader → AI → human
The AI becomes one of the invisible agents moving through the ordinary production of language.

Or , in a newspaper context, once AI is woven into translation, rewriting, social posts, editing, headline writing, research and so on, eventually asking "Was AI used?” becomes almost as useful as asking whether the writer used electricity.

The important question becomes:
What did it do?

The challenge ahead

Me:

But there is another problem for the humanist argument.

So far we've been asking whether AI can strip the human author out of a piece of writing without destroying its humanity.

Now we want to try the opposite.

Can AI put the human back in?

For more on this, see the next post in the Jamming series. But for now...

GPT (contd):

We've spent quite a lot of time establishing one proposition:
AI can remove the visible authorial self and still produce something recognisably human.
To Autumn was almost a laboratory experiment for this.
Keats deliberately suppresses himself. The objects do the work.
The Way Home pushed it further: remove the schoolgirl's elaborate interiority, and somehow the story became more emotionally resonant.
But the next question is:
What if AI can do the opposite?

What if, by absorbing an enormous quantity of human writing, AI becomes capable of constructing convincing representations of:
- grief;
- embarrassment;
- nostalgia;
- jealousy;
- religious longing;
- exile;
- sexual awkwardness;
- childhood;
- ageing;
- bereavement;
- political fear;
- the feeling of returning somewhere after decades away;
not because it has experienced them, but because language contains an extraordinary amount of information about how humans experience them?
That is a much more dangerous question for the traditional humanist argument.
Because then the answer to:
“But AI hasn't lived.”
becomes:
“No. But perhaps it has read almost everything we've written about living.”

And that doesn't establish that AI feels any of those things. It establishes something much more modest and much more interesting: it may be able to manipulate the linguistic representations through which humans recognise those things in one another.

Watching the wheels turn

GPT (contd)

That's the “linguistic machinery” idea.
Which brings us back to the angels.
The angels are not the source of the world they walk through. They don't replace the people. They don't explain the miracle.
They are another presence moving through the causal landscape, apparently invisible, but leaving footprints.
That's rather close to what you're describing AI as becoming in literature.
Not:
AI replaces the human writer.
Nor:
AI is merely a dumb tool like a typewriter.
Something much stranger:
AI becomes an active participant in the chain by which human experience is converted into language and then converted back into human experience in the reader.

Man versus machine is too crude.
The real object of study is the whole chain.

And that, incidentally, is why I think your phrase “medium of production” matters. A typewriter isn't neutral either. It changes what can be produced, how quickly, how comfortably, what gets revised, what gets discarded. The computer changed that again. AI changes it again. The reader changes the final event too.

I'd put it like this:
Biography is one ingredient in the production of literature, not necessarily its sovereign source.
Or, even better:
The writer is one participant in a much larger chain of production and reception.