How sci-fi writer JG Ballard’s computer poem predicted ChatGPT

How sci-fi writer JG Ballard’s computer poem predicted ChatGPT

In the 1970s, science fiction writer JG Ballard was intrigued by the increasing capabilities of computers – so used them to compose poetry. They are the first step on the journey to ChatGPT.

Novelist and short story writer JG Ballard, is best known for creating warped and reimagined versions of the world he inhabits. Dealing with the strange extremes of reality and often detailing the breakdown of social norms, his unconventional works are difficult to categorize.

Sitting on the edge of reality, these disturbing visions are often controversial. Eschewing far-future science fiction, Ballard described his own work as “a kind of visionary gift”.

Today, as we contemplate generative AI writing texts, composing music and creating art, Ballard’s insightful audience once again has something prescient and fresh to tell us.

In an interview from 2004, author Vanora Bennett suggested to Ballard that she write about “what will happen in a particular community”. Asked about “the kinds of real-life events” that inspired ideas in his fiction Ballard replied:

I just had a feeling in my bones: something odd was going on, and I explored it by writing a novel, by trying to find the unconscious logic that lay beneath the surface and find the hidden wiring. There seemed to be all these strange lights, and I looked for the wiring and the fuse box.

The topics in Ballard’s fiction often reveal how attuned he is to the subtleties of technological and social transitions that appear, as he says, beneath the surface. Society’s fuse box is often rewired in its ideas.

And with generative AI there’s definitely something odd going on, which Ballard’s attention seems to have been drawn to long before it happened.

As well as various OpenAI editions of the now famous ChatGPT, which generates custom text in response to simple prompts, there are various other applications emerging that automatically create cultural forms. Google Verse by Verse is an “AI-powered muse”, where you select a poet along with several criteria, such as number of syllables and type of poem, and it helps users complete the poem by generating lines in response to opening words entered into in the system. Sora is said to let you create videos from text commands. Different versions of Dall-E can convert text suggestions into visual artistic images. In the music field, apps like Aiva, Loudly and MuseNet can actively compose music on your behalf.

This is a snapshot of such a rapidly growing network of systems. They inevitably bring deep questions about human creativity and culture as we understand it. Nick Cave’s famous response to AI-penned song lyrics in his writing style was a strong and widely shared reaction to the perceived lack of an “inner being” behind the words. It, I thought, was just imitating creative thought. Others are now wondering if AI is putting an end to human writers.

As this debate continued, I discovered that a similar debate existed over 50 years ago. Looking through the archives of old art magazines that Ballard once edited, I discovered that he wrote about this futuristic concept in the 1960s, before continuing to experiment with the earliest forms of computer-generated poetry in the 1970s.

What I found was more than just revealing echoes of the past: Ballard’s vision actually revealed something new to us about this latest development in generative AI.

The surprise of computer-generated poetry
Listening to the audiobook version of Ballard’s autobiography Miracles of Life recently, a very short excerpt seemed to speak directly to this contemporary debate about generative artificial intelligence and the perceived power of large language models that create content in response to prompts. Ballard, who was born in 1930 and died in 2009, reflected on how, in the early 1970s, when he was prose editor at Ambit (a literary quarterly published from 1959 to April 2023) he became interested in computers who can write. :

I want more science in Ambit, because science is reshaping the world, and less poetry. After meeting Dr Christopher Evans, a psychologist who works at the National Physical Laboratory, I asked him to contribute to Ambit. We’re publishing a series of incredible computer-generated poems that Martin says are just as good as the real thing. I go further, they are things real.

Searching through back issues of Ambit from the 1970s I managed to find four items that appeared in the series Ballard referred to
Ballard says nothing more about this poem in the book, nor does he reflect on how it was received at the time. Searching through back issues of Ambit from the 1970s, I managed to find four items that appeared to be in the series Ballard was referring to. All appear to be computer generated and published between 1972 and 1977.

The first two are collections of what can be described as poems. In both cases, each small poem collected together has its own author’s name (more on that below), but the entire collection bears the authors’ names: Christopher Evans and Jackie Wilson (1972 and 1974). Ballard described Evans as a “hoodlum scientist” with “long black hair and a coral profile” who “races around his lab in a pair of American sneakers, jeans and a denim shirt open to reveal an iron cross on a gold chain”.

The 1972 collection was labeled with the overarching title “The Yellow Back Novels”, a play on the informal term used for popular fiction novels, and the 1974 collection was titled Machine Gun City. Both include brief notes that give more insight into how the poem was computer generated and Ballard’s thoughts on it.

The poem itself is, it must be said, a difficult read. I don’t want to speak for him, but reading the piece it becomes hard to believe that Ballard actually agrees with the assessment that they are “as good as the real thing” or, indeed, that they are the “real thing” – there may be an element of provocation in such a statement. However, quality aside, there is something interesting about how today’s debates about content generation – pushing us towards questions of what creativity is and what it means to be human – have precursors in these computer-generated works of the 1970s.

Ballard plots
Ballard’s view of the poems in 1974 seems to be consistent with recent comments included in his autobiography. A brief introductory note to the second collection opens with what it says is “the text of a letter from prose editor JG Ballard advising rejection of copies of famous writers”. Apparently, Ballard wrote the following, which is quoted parenthetically before the short cut:

B’s point is absolutely terrible – he’s absolutely dead-set and doesn’t seem to realize it … Even more interesting is this computer-generated stuff from Chris, which I think we should use some of. What’s interesting about this detective novel is that it was composed during a lecture Chris gave at a major psychology conference in Kyoto, Japan, with stories generated by an on-stage terminal linked by satellite to a computer in Cleveland, Ohio. Now that’s something this English experimental writer can think about.

Overseeing odd developments, he was drawn to new types of compositions
Whether these small computer-generated texts are stories, novels or poems is unclear and may be a secondary issue to the automatic production of culture on display here. Ballard seems to have taken with the new possibilities, and also seems to like the provocation given to other writers.

The image of the terminal on stage making a poem while its creator is busy speaking to the audience is a powerful image, replicated here by Ballard. He is clearly impressed by innovation and what it suggests about creativity. Overseeing odd developments, he was drawn to new types of compositions.

However, we probably shouldn’t take it at face value. The playful framing and anarchic tone warn us against being too literal. And there is another reason for us to be careful. Ballard’s interest may have been fueled by these events as he had written a short story featuring a machine that could perform the exact task of writing poetry some 11 years earlier. The short story itself seems to present a more questioning view of what computers mean for writing and creating prose.

Art imitates life
Written in 1961, Ballard’s Studio 5 story, The Stars featured an “avante-garde poetry review” editor working on the next issue. Sounds familiar. The poets he regularly edits all use an automatic “Verse Translator”, which they all refer to with established familiarity as VT. This VT machine automatically produces poems in response to set criteria. Poetry was perfected by these machines and so poets saw little reason in writing independently of their VT. When a poem is passed, hot from VT, the editor in the story didn’t feel the need to read it. He already knew that it would fit.

Poets had become accustomed to working with their VT machines, but their reliance on the machines for creative inspiration began to be unsettled by events. At one point the editor was asked what he thought was wrong with modern poetry. Although apparently a big fan of the automation of creativity, he wonders if the problem is “mainly a matter of inspiration”. He admits he “used to write a fair amount … years ago, but the impulse faded as soon as I could afford a VT set”.

Ballard’s story predicts that as creating poetry becomes a technical matter, the need to engage in the practice of writing disappears. In place of creativity, the editors suggest, is “technical mastery” which is “merely a question of pushing a button, choosing meter, rhyme, assonance on a dial, no sacrifice, no ideal to create sacrifice. worthwhile”. Not too far removed from the kind of prompts that today’s generative AI relies on to trigger its output. Often, as we saw with the examples of applications mentioned earlier, a set of criteria, phrases or any type of written instructions are used to initially direct the output of the generative AI.

A mysterious figure named Aurora, the story’s antagonist, declares disapprovingly that “they are not poets but mere mechanics”. When all the VT sets in the local area are destroyed by Aurora to “preserve a dying art”, the absence of human creativity is exposed. Not a single machine was left in one piece, even “Tony Sapphire’s 50-watt IBM was smashed to pieces and Raymond Mayo’s four new Philco Versomatics were smashed beyond hope of repair”.

In Ballard’s 1961 story, it is only the sudden absence of a working machine that prompts the poet to begin creatively rewriting
Editors are left with the next magazine issue to fill and no automated copy to fill it. Surprised by Aurora’s suggestion to “Write your own!” Tony, a fellow editor, offered some consolation, carefully reminding him that “Fifty years ago some people wrote poems, but nobody read them. Now nobody writes them either. The VT set just makes the whole process easier.”

In Ballard’s 1961 story, it is only the sudden absence of a working machine that prompts the poet to start writing creatively again. The dependency on VT was broken. The story closes with a burst of paper orders for three new VT sets. The story seems to be a warning against the automation of creativity and the implications it might have, should it arrive. In the 1970s it arrived in rudimentary form, and Ballard seemed, on the surface at least, to have a rather different reaction to its presence.

How do computers write poetry?
Each small section included in the 1972 and 1974 collections includes a title, author, and six lines of text. Those six lines are very formulaic. Some of that pattern can be seen just by squinting at many of the opening lines. These include such upheavals as “the thunder of the motor splits the lake”, “the roar of the jet shakes the house”, evokes “turbo fury crushes the crowd” and “Dr Zozoloenda stares as the plane glides”.

Although the 1974 pieces looked more diverse than the 1972 version, they kept the same type of formula. Part of the reason for the apparent consistency of form can be found in the brief endnotes by Evans and Wilson that close the first collection. They begin with the claim that:

This SF mini-novel has been produced by a computer programmed to write it, forever if necessary, given the command RUN JWSF.

RUN is a classic computer command to start a program. It’s not clear what JWSF stands for, but the vision is a perpetual writing machine that never stops and runs forever. They admit that the program itself is, as they say, “very easy”. They then go on to briefly outline how it works, pointing out that “a computer randomly selects from a specially selected group of keywords or phrases”. So this is randomly generated text from a curated pool of words.

There is also a structure where randomly selected words are placed. They explained that “the first line of the story basically consists of the computer completing the phrase: THE (BLANK) OF THE (BLANK) (BLANKED) THE (BLANK)”.

The use of random co-structures is presented as providing an almost endless source of new content

According to Evans and Wilson, in the structure of this opening line, “blanks are filled by searching through groups of words, thus ending with THE WINE OF THE MOTORS FRACTURED THE HOUSE or THE RUSH OF THE HELIOS HANGUS DESERT”. The two examples they give capture the feel of many of the opening lines in the short piece. The output is repetitive and predictable while also remaining strange. One mystery that remains is how word groups are created.

The use of random co-structures is presented as providing an almost endless source of new content that can be produced on demand. Evans and Wilson claim that their approach, “produces 10,000 possible unique sentences”. Following the opening sentence, they explain that “line two is a random selection of 10 complete sentences. Line three reverts to the strategy of line one. Line four is again a random selection of ten complete sentences, and so on”.

This interlaced structure of lines is central to all the pieces produced and published in 1972 and 1974. The second mystery is how the randomly selected complete sentences for the interlaced lines were produced and chosen. There don’t seem to be any other effects, so these details will likely remain unknown.

The generation of permanent material is first framed in terms of the number of sentences possible. But something more reflective is also introduced, which is the generation of ideas. Evans and Wilson asked themselves: “How many original and unique SF mini-novels can a computer generate before running out of ideas?”

Their seemingly speculative answer, since we don’t know how many words are included in the group, is simply “typing at a rate of 10 characters per second, this would take (approximately) 10,000,000,000,000,000,000 [one hundred quintillion, or 10 to the 20th Power] years that will probably see the Universe come and go several times”. The generation of poetry by this machine is, in other words, without any real limits. Obviously, we can question whether it actually had any “ideas” in the first place.

Like the content, the author’s name attached to each six-line work is also computer-generated. The author’s name is again “selected from an appropriate pool of SF-type names, paired in the same random fashion”. What constitutes a SF genre name is not clearly explained, but some of the author names that have come up are things like ZQ Johnson, Blade Sinatra, Frank Archer, Marsha Fantoni, Blade Van Vargon and even Tagon “X”.

Adding a name humanizes the writing in some ways, even if the name itself mostly seems to be clearly made up. Giving this work a named author actually draws attention to the question of authorship and the intersection of human creators with technology.

Computer-generated poetry or hoax?
Subsequent articles published in Ambit in 1976 and 1977 did not seem to live up to the promise made in 1974 that these small pieces would appear in “an endless stream in Ambit”. Instead, they changed direction somewhat, moving from computers creating text to interacting with humans. A 1976 piece entitled Hallo, your computer calls, again credited to Chris Evans and Jackie Wilson, presents a strange if prophetic interaction introduced as “an experiment to see if computers can help doctors diagnose disease”.

A 1977 work “The Invisible Years” is more puzzling. This time credited to Tim Bax, JG Ballard, Chris Evans and Ronald Sandford, the work is presented in an awkward corner box and described with an opening statement: “This year Ballard answers the questions of Chris Evans and the computer. For a painting inspired by Mr. Ronald Sandford. ” The strange intervention appears to have been the final installment in this series of computer-generated contributions.

We may begin to question, especially with its strange framing and increasingly strange content, whether this is actually a computer-generated text at all, or, given the type of publication and those involved, if this is another form of expression, perhaps parody or satire even . The chapter in Ballard’s Wonder of Life where Evans is discussed shows that their collaboration spread into fictional ideas as well. It may also be a kind of hoax, designed to raise questions about what automation means for culture and ideas.

Enthusiasm for the possibilities of computer writing was evident even then
It is now impossible to confirm what exactly happened or what, if any, technology is being used. It appears that a computer program was involved in some way with the final product, and whether it was a fully automatic writing bit is actually a side issue when considering consider the importance of this small work. Whatever we see in this strange automatic poem, this case reveals something about the kind of interest in idea generation and computational cognition that is at play in more advanced forms today. Here from the 1970s shows how this logic has evolved.

Enthusiasm for the possibilities of writing computers was evident even then. Yet the apparent passion attached to Ambit’s poetry may also be a reaction, or even an ironic and playful reaction to, the emerging computer systems and even the burgeoning AI of the 1960s and 70s. The questions surrounding creativity and human values implicit in Ballard’s short stories may hint at this. But the kinds of questions, results, and implications of computer-generated writing have not yet condensed into the kinds of debates we see raging today.

Sensitivity to automatic creativity
If we take at face value the description of the generative process described in the notes accompanying this poem, as well as mentions in Ballard’s later autobiographical account, then the main difference between the small pieces Ballard commissioned and today’s popular turn to AI is the shift from randomness to probability. Generating a poem that draws randomly on a curated group of texts is quite different from generating text based on probabilistic calculations from a large data set. Yet the underlying sensibility and logic are the same, both informed and driven by a changing desire to automate more aspects of social and cultural life.

Whether what we see with this 1970s poem is genuine or if it is a kind of performance or playful satire, it still reveals something about the emerging attitude towards the possibilities of computational creativity in its early form.

Ballard’s enthusiastic response to the new possibilities suggested by poetry in the 1970s contrasts with the more dystopian vision of his 1961 short story. Ballard seems to embody what I call the tension of algorithmic thinking – by which I mean the unresolved and competing forces that push simultaneously in different directions as we confront ever-increasing automation. On the one hand we have the problem of the removal of humans from human activity, on the other hand we have the removal of knowledge from the creation of culture. The short stories and poems in Ambit both capture the tension that accompanies today’s AI-generated text, art and music.

We may be shown from a different perspective, to use Ballard’s own phrase, the wiring box and fuse of creativity. Ballard’s attention was focused on “something odd happening”. That peculiarity deepens as the use and application of generative AI continues to grow.

About Kepala Bergetar

Kepala Bergetar Kbergetar Live dfm2u Melayu Tonton dan Download Video Drama, Rindu Awak Separuh Nyawa, Pencuri Movie, Layan Drama Online.

Leave a Reply

Your email address will not be published. Required fields are marked *