Franz Kafka once wrote a short story entitled A Report to an Academy, in which an ape attains human intelligence and gives a lecture on its past as a wild animal.
Something similar happens when you ask ChatGPT about its own limitations: “Are you capable of writing a good poem?” write, most poetry that is considered relevant and meaningful is usually penned by a poet with a distinctive identity and voice.”
The technique recognizes – with great composure – that the work of an author without biographical interest usually does not attract much attention from the reader.
Since it launched a few months ago, ChatGPT could become the most prolific poet in literary history. Thanks to immense neural deep learning networks, the technology is able to write texts in a few seconds, based on very simple instructions and without plagiarizing a single sentence from the Internet. However, the AI cannot legally be considered the author of these works.
“Maybe a machine can create art… but it’s not protected by intellectual property law because there has to be an author for it. As such – except in rare cases – only [human beings] can be authors,” says Mario Sol Muntañola, lawyer and intellectual property expert.
Guillermo Marco and Julio Gonzalo – natural language processing researchers at Spain’s National Distance Education University – have spent years exploring the limitations and possibilities of large language models like ChatGPT. They declare themselves skeptical about “the hype” surrounding this technology.
“These models work like networks of artificial neurons that are similar to our brain. They learn to write by cognitively simulating our reading,” explains Marco, co-author of the poetry collection Other Clouds, for which he received a poetry prize in 2019.
In one of their first experiments, Marco and Gonzalo asked a group of participants to rate six aspects of the synopses of books and films that this artificial intelligence produced.
“We gave [each product] made up a name and the machine created a storyline for that title,” explains Gonzalo. The result was that the language model outperformed humans in every area… except creativity. The researchers then decided to focus on that specific aspect and changed the aim of the study.
“We started experimenting with poetry because synopsis—at least for humans—isn’t inherently creative writing,” the men note.
Before measuring the results, they suggested extracting a definition of what users understand by creativity.
“It’s different for everyone… but almost everyone agreed that creativity is the original, the unusual – what they’ve never seen before,” explains Marco.
The French poet Philippe Soupault is considered one of the fathers of “automatic poetry”. Marc GANTIER (Gamma-Rapho via Getty Images)
The duo concludes that these language models are not particularly creatively designed.
“They are meant to imitate rather than be original. They learn to say the least surprising,” they affirm.
Marco explains this with an example: “If you give [the AI] a sequence – “Heaven is…” – and you ask them to complete the sentence of 50,000 words [the technology] learned, it will always choose the word “blue” because it is the most common.”
ChatGPT’s priority, they explain, is preserving the meaning of the text — making it difficult to make a sentence aesthetically pleasing.
“It turns out there are people who ask the machine questions that already have such a high level of originality [the AI] has no choice but to improvise,” Gonzalo clarifies. “I read an example where someone asked ChatGPT to write a Bible-style story about a person who put a sandwich in the slot of a VHS player. The result was hilarious.”
It’s also important to note that ChatGPT doesn’t understand the words it’s learning.
“All [its] Knowledge is intuitive – by reading it learns what language is. But [the AI] does not have the ability to reflect or think rationally. This is precisely why it learns to imitate sonnets, but is unable to explicitly recognize that there is a rule relating a certain type of rhyme to a certain number of verses,” adds Gonzalo.
Marco emphasizes that these language models will always be limited to the input sequence, i.e. to the instructions of a human.
“The [large language models] will never have an original idea, they will never have an intention. That intention will always come from a human being in the moment. What would be truly artistic is for [the AI] to refuse to write or to choose to write [the prompt] in his own way.”
In his book Uncreative Writing: Managing Language in the Digital Age (Columbia University Press), Kenneth Goldsmith writes: “Perhaps the great writers of the future will be those who can write the best programs to manipulate the practices of language, too to analyze and disseminate.” This opens up the possibility of literature becoming a collaboration between machine and human—a process designed not to replace human creativity, but to enhance it.
The writer Jorge Carrión – who has published numerous papers on ChatGPT-2 and 3 – has analyzed The Magnetic Fields, a book of poetry by André Breton and Philippe Soupault. Published in 1919, it is considered the birth of automatic poetry – writing without conscious control or censorship. Carrión compares the emergence and influence of the surrealist literary movement in the first decades of the 20th century to the spread of large models of language.
“If the transition between conscious and unconscious writing characterized these years, the writing produced by machine learning and other forms of artificial intelligence gives ours a special atmosphere.”
Speaking to EL PAÍS, he says that “algorithms write very well – they write almost perfectly, they access areas that are taboo for humans – but they are still incapable of brilliance, of metaphor, of knowing the best poetry. That doesn’t mean they don’t write better than many poets, [like] poet-influencer.”
However, he does not hesitate to say that the arrival of a technology capable of writing good literature is only a matter of time.
“The intelligence of algorithms, robots, neural networks, or programs or formulas that artificial intelligence does not yet have … may end up being metaphor and irony in new and undoubtedly literary forms.”
Sign up for our weekly newsletter to receive more English language news from the EL PAÍS USA Edition