With the turn of the century artificial intelligence has emerged out of the realm of science fiction and has found various practical applications in the real world, like self driving cars, FDA approved AI software for detection of eye diseases, sorting resumes, adding to your smartphone camera’s creativity, and even the Pentagon taking note of the possibility of artificial intelligence arms race, that’s after beating humans in chess and Jeopardy! (a language based game). AI can master problem solving that involves logic, mathematics and algorithm, it builds on the existing knowledge and learns to identify solutions, AI finds solutions to complex problems which would be daunting for unaided humans. Can AI also replicate humans in the sphere of creativity, can AI acquire creative strength as it evolves, or is creativity reserved solely for highly evolved sentient beings like humans?
Paradox
The arguments are ample on both sides. Through machine learning AI can recognize faces, interpret languages, write compelling ad copies and email subjects and beat humans in chess. Would you call this creative work? What about evoking human emotions through stories based on experiences and emotions, can robot writers do that? Machines can understand and reproduce patterns of storyline with different possible twists, but do they understand the depth of the story, unless the information is fed to it by humans? Machines can answer complex problems but can they ask questions?
In 2016 a Japanese AI software made an attempt to co-write a novella with human writers and almost won the Hoshi Shinichi Literary Award. The story was titled “The Day a Computer Writes a Novel”. “I was surprised at the work because it was a well-structured novel,” said science fiction writer and award judge Satoshi Hase. Jacob Brogan, an editor at the Washington Post was but cautious before getting impressed about a robot writing a story, the part that the novel was being co authored by humans wasn’t an excitement for him. Upon close examination of the event we get to know that the AI in question only remixed a novella already written by humans, the book’s components were broken into words, sentences, and basic structure before being fed to the machine.
Computers work on logic and math and theoretically there is no limit to how much they can learn, if they can learn to play chess they can be taught to learn what sort of writing humans like to read, and eventually may produce a best seller, they might even learn to paint. While a human writer is inspired by real life experiences and emotions, a robot is not capable of the same (as of now). We always have the writer of the story in the back our mind and unconsciously relate to his experiences or thoughts while moving through the story, how can we imagine the same for a robot?
The question is how would humans respond to this artificial creativity… Is there something more than raw creativity that humans find innately alluring… something that emerges from human experiences and emotions and can be connected to our lives rather than a robot’s life. Do we always need a human touch to a piece of art… it’s a paradox.
The Sophisticated Evil
Elon Musk’s non-profit OpenAI has developed and recently released GPT-2, a large language model that can generate realistic paragraphs of text. GPT-2 can generate coherent text when given a snippet of input. The tool’s stunning efficient made it vulnerable to misuse and hence was released in stages and with caution. In its release blog OpenAI states that humans find GPT-2’s output convincing but also the programme can be fine-tuned for misuse. Fake news is on the top of the list of possible misuse which contains impersonating others online, automating the production of abusive or fake content for social media, and automating the creation of spam and phishing content. In 2016, Microsoft’s chatbot Tay, supposed to emulate a teenage girl online, within few hours of its release was fine tuned by somebody which resulted in the chatbot calling feminism a disease, approving of Hitler’s actions, abusing Barack Obama and spewing more hateful tweets.
Note: Elon Musk advises mankind to be extremely cautious about artificial intelligence, he has called AI humanity’s ‘biggest existential threat’ and compared it to ‘summoning the demon.’
Applications with Limited Creativity
While human’s creativity may not remain unrivalled in future, machine learning so far is best at understanding stories with a typical format like, quarterly reports of companies, weather or sports snippets. Salesforce’s Einstein Copy Insights uses machine learning to enable marketers to identify optimal words for email subject lines, text messages, social media copy and similar text. It also notifies users about under-performing marketing emails and offers recommendations for improvement. Google Search uses natural language processing to better understand how different words relate to each other in a search query. Microsoft is developing software to generate text for ads and Wall Street Journal is using ai to enhance its news coverage.
Heliograf is Washington Post’s robo reporter which churned out 300 reports during the Rio Olympics in 2016 and it helped the post cover House, Senate and gubernatorial races for all 50 states on election day. Currently this type of tools can be used to automate a bulk of news writing in some segments which constitute a large section of news, saving time for journalists or resources for publishers. The Associated Press has used bots to cover financial stories and USA Today has used video software to create short videos. Google’s Perspective API has been used by the New York Times and Disqus to moderate comment section on their news articles and blogs
Scholarly Writing
The most interesting use of artificial intelligence in content creation happened in scholarly publishing recently. Springer Nature published its first AI generated research book titled ‘Lithium-Ion Batteries: A Machine-Generated Summary of Current Research’. The pdf file, which is free for download mentions that the book is ‘cross-corpora auto-summarization of current texts from Springer Nature’s content platform “SpringerLink”, organized by means of a similarity-based clustering routine in coherent chapters and sections’. It automatically condenses a large set of papers into a reasonably short book. Henning Schoenenberger, Director Product Data & Metadata Management at Springer Nature, points out that this method allows for readers to speed up the literature digestion process of a given field of research instead of reading through hundreds of published articles. The book was presented at the recently concluded Frankfurt Book Fair 2019, here’s a video if you know Deutsch!
When asked if there’s a scalable business model in machine-generated book publishing Henning Schoenenberger said, ‘I do expect that machine-generated content will become a scalable business model at some point. However, as with many technological innovations, we also acknowledge that machine-generated research texts may become an entirely new kind of content with specific features not yet fully foreseeable but we assume that in future there will be a wide range of options to create content – from entirely human-created content, a variety of blended man-machine text generation to entirely machine-generated text.’
Here’s a video of a song created by text predictor mixing words from Bob Dylan and restaurant reviews in New York. It says ‘if you wanna travel underneath the worst bar in the West, knock on this establishment again’. Quite Impressive!
