It would be easy to be anxious, if you worked in the publishing industry by stories of how AI is transforming the world – and, like many, to be considering how your own role will be affected. Feelings may well run deep, given the impact that digital print has already had on areas of publishing.
Progress in digitization has been limited over the years, to the speed of internet connections. For example, it wasn’t until home fixed broadband connections became available for the majority that YouTube took over.
It was only when mobile bandwidth went 4G that people started watching films on their mobile devices. Unfortunately, publishing or the use of the written word was one of the first industries to be disrupted. One useful definition of which, in my view is that disruption means its business model fundamentally changed. This change caused a drop off in print journalism but also, on the other side of the coin, a rise in the distribution of books in the form of audio books and LED readers like Kindle.
Now, here comes another aspect of technology, AI, which could disrupt things again.
First of all, while AI doesn’t ‘understand’ things in the sense that humans do – and all estimations are that this consciousness is likely at least 20 years away, and might involve a completely new starting point from where the research is now, AI has already made forays in to print media, starting with the more rote aspects of journalism.
Layoffs off of reporters and editors are being seen first at publishers like Reuters and Bloomberg News, thanks to their automated AI technology. Cyborg, as it is named, accounts for an estimated one-third of the content published by Bloomberg News. It is able to examine financial reports as soon as they are available, and create news stories featuring the most pertinent information. Best of all, it can do this faster and more accurately than a reporter – who would usually find this type of work sleep-inducing.
AI robots have not been limited to financial reporting either. The Associated Press and The Washington Post have been using AI to produce articles for minor league baseball and high school football respectfully. Los Angeles Times have reported using robot reporters to write about earthquakes.
Undoubtedly, we are seeing the endemic rise in the use of artificial intelligence; however, journalism executives are quick to point out that this doesn’t spell an end to human journalism. It is suggested by publishers leading the way into this new era of publishing, that AI will simply allow journalists to spend more time on more practical work, and that artificial intelligence should be seen as merely a part of the toolbox.
Lisa Gibbs, director of news partnerships for The Associated Press, says, “The work of journalism is creative, it’s about curiosity, it’s about storytelling, it’s about digging and holding governments accountable, it’s critical thinking, it’s judgment — and that is where we want our journalists spending their energy.”
While AI is on the rise in many industries, the field of journalism has been largely shielded by the fact that machines have long struggled with the intricacies of language. Even with recent advancements in deep learning powered by big data and better processors, AI has failed to solve this complex intellectual challenge.
But now, new systems of analyzing text from researchers and large corporations like Google and Open AI, are edging closer to a significant breakthrough.
Open AI’s algorithm, called GPT-2, is currently the most extraordinary example. It is superior at a task referred to as language modelling, which tests a program’s capacity to guess the next words in a sentence. The ability of the algorithm is truly mind-boggling. Give GPT-2 a headline, and it can write the remainder of the article, complete with bogus quotes and statistics.
Recently, Open AI rather dramatically withheld the release of their newest language modeling algorithm, GPT-2 – instead deciding to release a small, simplified version of GPT-2 with its sampling code and research paper for researchers to conduct further experimentation.
Open AI feared that the full release of GPT-2 could see it being used to automate the mass production of misinformation. The decision also accelerated the AI community’s ongoing discussion about how to detect this kind of fake news. In a new experiment, researchers at the MIT-IBM Watson AI Lab and Harvard NLP considered whether the same language models that can write such convincing prose can also spot other model-generated passages.
The idea behind this hypothesis is simple: language models produce sentences by predicting the next word in a sequence of text. So if they can easily predict most of the words in a given passage, it was likely written by one of their own.
The researchers tested their idea by building an interactive tool based on the downgraded, publicly-accessible version of Open AI’s GPT-2. When you feed the machine a passage of text, it highlights the words in green, yellow, or red based on their decreasing ease of predictability; it highlights them in purple if it wouldn’t have predicted them at all. In theory, the higher the fraction of red and purple words, the higher the chance the passage was written by a human. Vice versa, the higher the share of green and yellow words, the more likely it was written by a language model.
Ultimately, it could be argued that publishing is merely a microcosm of the impact that AI is having on every industry. The ‘State Of The Art’ with AI technology, Machine Learning, can only, at this point, serve to aid productivity of journalists and writers by completing sentences and performing plagiarism checks, to see what proportion, for example, of an article, has been written by an AI.
More broadly, the benefits of AI which apply to every industry also apply to publishing. Improved analytics to help them sell more, more personalized delivery of the content each individual wants to read.
The challenge, for publishing, is to accept the tsunami of change, acknowledging that this is the second major change they’ve dealt with – after the initial digital disruption. Publishers need to focus on how best to reap the benefits of what AI offers, without feeling nervous or being perturbed by the potentially negative consequences.
AI cannot now, nor will it in any currently foreseeable way ever replace the real purpose of publishing – to transmit ideas, stories, wisdom and inspiration between humans.
After all, only humans have life experience, imagination, and can tell stories that others want to hear. Only humans have the intuition to form a sentence which communicates the essence of something in a way that connects with other human beings.
So long as publishers are looking for written work which leverages those innately and uniquely human traits, it would be safe to assume that the industry will always require written works authored by humans.
Ralf Llanasas is a technology blogger, covering the latest mobile phones and technology developments. He currently works at Whatphone.com.au.