AI Has Come to Save the Arts from Themselves

Comment on this story

Comment

In 1959 – or the year 63 before ChatGPT – Soviet poet Boris Slutsky wrote “Physicists and Lyrics”, a poem(1) asserting the defeat of poets by engineers:

It’s not demeaning either

Actually, it’s pretty fun

Solemnly in logarithms.

Since modern generative AI can actually produce poetry that a published verse writer can admit is sometimes better than his own, and generate images that a millionaire digital artist can unquestioningly praise, verse of Slutsky may be considered unintentionally prophetic. But he also wrote “Lyrics and Physicists,” another poem on the subject, just a few years after the first:

You’ve only won one battle,

You haven’t won the war yet.

We have barely drawn our swords

Halfway out of their sheath,

And as the saber is slipping free,

By illuminating the arm muscles,

Poets are not offended with you –

Herein lies the true prophecy. The engineer’s disdain for the direction of the humanities has culminated in the creation of DALL-E and ChatGPT just in time to save the arts and humanities from themselves. Salvation, however, is likely to come at a high human cost.

The technological revolution that began with personal computing has made creation so easy that moving into a “soft” profession is the path of least resistance these days. No longer does a writer need to spend months in dusty archives researching a book—or days sifting through print materials to produce a column. No middleman is required to publish your writing either – you can go directly to the audience. Visual artists, for their part, have mastered digital tools so perfectly that good technique is less of a requirement than ever before; photography, illustration and animation have become commodities. Only the latest advances in generative AI are beginning to threaten the livelihoods of hundreds of thousands of professionals in these fields. But this is a development that should have been expected.

True, those of us in the creative trades—and that includes me, my wife, and our oldest daughter (they’re both artists)—have long made less money, as a group, than engineers. But less and less effort was required of us to produce what we sold. We’ve leisurely Googled, used electronic editing tools, online thesauruses and translation, created artwork using commercial software. It was inevitable that at some point, the tools would overtake us and take over some of the tasks the market was sending us.

I can understand the anxiety of the average illustrator – and yes, Beeple’s too – as they look at a Midjourney creation. I get the fear in the throat of a columnist who has fed an idea to ChatGPT and received a text that would pass the editor with minimal changes. They – we – were taught from childhood that their creativity was uniquely theirs. The fact that a machine learning model can reproduce it if fed enough training material is a shock. The first reactions are remarkably human: Outrage that the models use people’s actual work as training data without respecting copyright, a search for ways to trick the software into making mistakes or saying something politically incorrect. Here you go! At least humans will be needed as a sort of moral police for AI robots!

My wife has laughed at the way AI “artists” inevitably misuse hands, paws and feet: The number and shape of the fingers are – look – unpredictable. An artist who lacks technique will hide his hands in his pockets and his feet in the grass. I’ve scoffed at ChatGPT’s inane, timid writing and its audacity when it’s absurdly wrong. We both know, however, that for most consumers and for most commercial uses, what AI can do is already enough; for a manager on a budget, it is virtually indistinguishable from what a human would have been hired to create a year or two ago. As for the bugs, they will be fixed, as they always have been.

It’s futile to ask for more regulation, more copyright protection, tighter content control. The weakest among us will, of course, find sustenance in these forms of political control, but this will be a cop-out: if one believes in freedom of expression and agrees that “good artists copy, great artists steal,” then everything. in the creative sphere that applies to humans should also go to robots.

The time has come to measure ourselves against higher standards, above all those of originality and creative prowess. The key feature of modern AI is that it is only as good as the training data. If what you do is new and truly original, you have no reason to fear: By the time AI arrives, you’ll be gone. If you challenge convention, it will be difficult to imitate based on that convention – and all the checks and balances created by AI creators to avoid shutdown will be in your hands.

It is also a time of reckoning for people of modest talent. The modern education system has made sure that they are not left behind and has convinced many of them that it is enough to be “uniquely you”. But much that passes for writing or art these days might as well have been produced by a deep learning model—it has taken more from the cultural context than it has contributed to it.

True genius, on the other hand, will beat an algorithm every time – even if its product is in the public domain and thus part of the training data.

The oak that in summer was sweet to hear, And rustled its leaves in the autumn of the year, And whistled and roared only in winter, is gone, And the birch has grown in its place. – The knight’s bones are dust, And his good sword rusts; I believe his spirit is with the saints.

This is from Coleridge’s The Knight’s Tomb. I asked ChatGPT to produce a poem about a knight’s grave, an oak and a birch in the style of Coleridge. It came back with a string of verses that ended like this:

The oak and the birch, stand side by side Two guardians of the forest, with intertwined roots, guard the knight’s grave, through sun and rain, until time itself has forgotten his name

And so the knight stands, his tale already finished Beneath the oak and the birch, in the heartbeat of the forest, his memory lives, in the rustling of the leaves A memory of the past, that never leaves.

The terrible banality of this production lies not only in the unnecessary, irrelevant words, nor in the logical contradiction between the last verses: It is in the simple idea that the two trees “guard” the grave together. The bot poem tells the wrong story because its creator is mediocre by design.

The challenge of living in a world where generative AI exists and is constantly improving its capabilities is to discern whether the difference between your work and that of a machine is as stark as it is in Coleridge’s example. If so, chances are enough people will appreciate your work for you to have a comfortable existence.

It still takes guts – like all good digital tools in the old days – to stay in the creative trade. It also takes courage to admit that software can put you out of a job—and, by implication, that you’re in the wrong job. It’s a painful admission to make, but even more painful to avoid: The financial consequences can be devastating. So if you’re just a varsity and not a Coleridge, take Slutsky’s first poem to heart – and do your best to join the winning side.

More from Bloomberg Opinion:

• Did a robot write this? We need watermark to find AI: Parmy Olson

• ChatGPT could make democracy even messier: Tyler Cowen

• Musk should keep Twitter’s most vital feature: Gearoid Reidy

(1) Translations by the author. The original rhymes and meter are not preserved

This column does not necessarily reflect the opinion of the editorial board or of Bloomberg LP and its owners.

Leonid Bershidsky, formerly a Bloomberg Opinion Europe columnist, is a member of the Bloomberg News Automation Team. He recently published Russian translations of George Orwell’s 1984 and Franz Kafka’s The Trial.

More stories like this are available at bloomberg.com/opinion