Last year, The Economist organised the Open Future international essay competition on the subject of Climate Change. Nearly 2,400 youths from over 110 countries entered the contest to answer the question, “How to respond to climate change?”
Intelligence enthusiasts at the British weekly got thinking. They wanted to know how artificial intelligence (AI) would respond to the same question. They decided to find out. They fed the essay question and a 58-word description through GPT-2, the natural-language processing (NLP) algorithm that has taken the world of AI by storm ever since its partial release in February 2019 by Open AI, the Elon Musk-funded research organization in the US.
Judges were asked to evaluate the AI-written essay using the same standards they used with hundreds of competing essays, but without telling them it was the handiwork of machine or artificial intelligence, and not the human brain. The essay began by asking: “Do we want to go through the same process we have been through for decades with no changes? Is there a way to build a sustainable energy system that is both affordable and environmentally responsible?” To think, these accusatory responses come from a machine, and not the human brain!
The fact? GPT-2 is damn good at writing essays. The ultra-sophisticated text-generator, which was deemed “too dangerous to release,” can write full, coherent sentences and paragraphs; in fact, an entire fake news article in a matter of seconds. Feed a word or a sentence to the neural network and it will do the rest for you, things you always wanted to write about but never did because you could not guess your own thoughts. It is that freaking incredible!
The story of English-speaking unicorns is now stuff that legends are made of. Here’s the sample text, pure that took internet by storm:
Human prompt: In a shocking finding, scientist discovered a herd of unicorns living in a remote, previously unexplored valley, in the Andes Mountains. Even more surprising to the researchers was the fact that the unicorns spoke perfect English.
Text generated by the GPT-2 algorithm: The scientist named the population, after their distinctive horn, Ovid’s Unicorn. These four-horned, silver-white unicorns were previously unknown to science. Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally solved. Dr. Jorge Pérez, an evolutionary biologist from the University of La Paz, and several companions, were exploring the Andes Mountains when they found a small valley with no other animals or humans. Pérez noticed that the valley had what appeared to be a natural fountain, surrounded by two peaks of rock and silver snow. Pérez and the others then ventured further into the valley. “By the time we reached the top of one peak, the water looked blue, with some crystals on top,” said Pérez.
Little wonder that GPT-2 is billed as “humanity’s greatest text generator.”
With high levels of sophistication in deep learning, the publishing industry is surely headed for interesting times. As it is, technological advancements, especially in the field of artificial intelligence, have taken great strides in the business of publishing, breathing into the sector much-needed fresh air. Technology-enabled tools and analytics have come in handy at all stages of publishing.
If you are a news publisher, an intelligent machine can replace an army of journalists used in covering routine events, with greater speed and accuracy. Automated journalism is already a reality.
If automated text analysis helps detect plagiarised passages, AI-enabled programs allow writers and editors to self-edit their work, thus reducing editorial workload. From management of contacts, rights and royalties and image recognition to auto text tagging and automated formatting, AI has solutions to all kinds of publishing needs. Particularly for new writers and self-publishers, AI is hugely empowering.
“Al-enabled analytics can identify and adapt to consumer search trends and predict the ‘next big thing’ with more accuracy than traditional market research methods,” Max Kalmykov, VP, Media & Entertainment at DataArt, wrote on hackernoon.com.
While creativity was never traditionally associated with artificial intelligence, neural networks have done poetry and even novels. The first machine-written novel, 1 the Road, came out in 2018. It was a literary travel fiction, à la Jac Kerouac, written by Ross Goodwin’s computer after a cross-country drive. A former ghost-writer of the Obama administration, Goodwin got his algorithms to convert CIA torture reports into a novel in 2014. He uses his neural networks to generate poetry and screenplays too.
Author Sigal Samuel is using GPT-2 to help her write her novel by “knocking my imagination off-kilter a bit, nudging it in more surprising and playful directions.”
The judges at The Economist were asked to rate the essays as “yes, maybe or no”. Only two of the six judges marked them “maybe”. Interestingly, none of them marked them “yes”. The AI-written essay was okay enough to make sense, but did not make the cut. The key perhaps lies with emotional intelligence, the unique human trait that distinguishes man from animals and machines. Writing is a creative process that is fuelled by human emotions. An intelligent machine generates text, an emotionally intelligent human brain creates, it writes. It takes into consideration not just raw data but various tangible and intangible inputs that the human brain is able to process. Machine learning can come out with text — word by word, sentence by sentence and chapter by chapter — that resembles the work of the human brain. But that’s only on the surface; since it cannot emulate emotional intelligence, the writing is devoid of human logic or empathy.
However, machines are trying to get there too; emotional intelligence is said to be the next frontier for AI. Some machine learning experts actually believe AI can get as emotionally intelligent as humans, if not more. What that would mean for the publishing industry is hard to visualize given that intelligent automation is totally redefining the rules already. Let’s hope that at least the end reader or content consumer is still the king, and a human being.