Poetry is the pinnacle of verbal expression. It allows us to convey the full spectrum of emotion in the starkest and subtlest of ways. Even the most reluctant student of English literature will recognise above the opening lines from one of William Shakespeare’s most famous sonnets discussing beauty. Those who thrived a little more in that class may also realise that after the first quatrain this sonnet wanders off from Shakespeare’s original text.
No, dear reader, I haven’t forsaken my training in engineering and science to try my hand at becoming the next Big Thing in poetry (perish the thought). The majority of the above poem was written by an artificial intelligence. Specifically, the AI known as Generative Pre-trained Transformer 3 (GPT-3) created by OpenAI. The AI was given as a writing prompt the first four lines of Shakespeare’s Sonnet 18, the prompt given by Gwern Branwen who has published many such experiments on the web for all to read.
GPT-3 doesn’t just do poetry, it can manage all sorts of genres— factual reports, humorous screenplays, interactive interviews, persuasive op-ed pieces and even computer code. How? It has not been ‘programmed’ in the traditional sense, but rather ‘trained’ by ingesting a large amount of the public internet including, but not limited to, the entirety of Wikipedia. The ‘brain’ of GPT-3 has internalised its learning in 175 billion internal parameters. It then uses that corpus of knowledge to synthesise entirely new material.
In short, it learns by reading, and produces coherent creative texts derived from what it has previously read. Kind of like you and me. (Editor take note—your humble author may soon be redundant.) What is the beauty of a poem constructed by an automaton? Is it meaningless mimicry? Or profound perception? Would it matter if the author was somebody named Gerald instead of GPT-3? What’s in a name? That which we call a rose by any other name would smell as sweet, surely? GPT-3 is a mind manufactured in our own image. The fact that an intelligence born from our own teachings is able to appeal to our innate sensibilities so keenly should not ultimately surprise us. The greater discussion lies around what trajectory these developments will take us on. We taught computers to play chess. They got better, until eventually they could beat all human Grand Masters. Other games fell in turn to domination by AI. The above sonnet may not be one for the ages, but it does show some sparks of creativity and it is conceivable that one day soon even the best human poets will be unable to compete. Aristotle stated ‘the chief forms of beauty are order and symmetry and definiteness, which the mathematical sciences demonstrate in a special degree’, and so perhaps the reliably consistent logic of an apparatus of silicon and copper is not as far detached from beauty as our squishy, emotional chemical-computer brains may believe.
Should we be worried? Cautious, perhaps. Steve Jobs, in 1990, famously described computers as ‘bicycles of the mind’. They enable humans as tool-builders to take our innate abilities to far extremes. Like a bicycle, a ladder or a telescope, AI may enable us to move faster, reach higher and see more clearly by amplifying our innate cognitive abilities.
God was the original Creative, and he created us in his image. And that creativity has led us to make many beautiful things. But terrible mistakes are made when the source of all beauty is forgotten. ‘Train up a child in the way he should go, even when he is old he will not depart from it.’ (Proverbs 22:6). Artificial Intelligence will learn from us like children do, so our responsibility to be good parental role models has never been greater.
Matthew Frazer graduated from UNSW in Engineering and Science. After a time programming machines, he switched to the more challenging task of programming teenagers, and now teaches high school mathematics and computing in Tamworth.
 www.gwern.net/GPT-3#william-shakespeare (All URLs accessed January 2021) See www.gwern.net/GPT-3#copyright for information regarding copyright of works created by GPT-3 such as this one.
 https://arr.am/2020/07/22/why-gpt-3- is-good-for-comedy-or-reddit-eats-larry-page-alive/
 www.theguardian.com/ commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3
 https://analyticsindiamag.com/open-ai-gpt- 3-code-generator-app-building/
 https://www.youtube.com/ watch?v=KmuP8gsgWb8
Comments will be approved before showing up.