Artificial intelligence is one of the year’s most talked about subjects. The increasing penetration of voice search and home assistants, and the adoption of chatbots as a part of the marketing mix has prompted marketers to think about how they can leverage AI to create effective content. In a recent World Advertising Research Centre survey of marketers, AI was cited as the most important emerging technology for their business in 2018.
The launch of IBM’s Watson Advertising, with an AI application to create content, has certainly fuelled expectations. But what will the future look like? Will people be targeted with highly personalised messaging based on facial recognition, as seen in the movie ‘Minority Report’? And will machines be able to replace creatives to write content?
Visual and speech recognition technology has improved significantly in the last few years which could enable much more targeted messaging, but marketers must be careful not to cross the line between clever and creepy or intrusive. Even though machines are good at delivering pre-programmed or pre-learned content based on facts, their ability to create content from scratch is still limited, and the content lacks emotion. Recent attempts at using AI for creative work have included attempts at writing new novels. First reported in The Verge, Botnik Studios trained an algorithmic tool on J.K. Rowling’s seven Harry Potter books and then used predictive text AI to write a three-page chapter. The chapter produced was titled, ‘Harry Potter and the Portrait of What Looked Like a Large Pile of Ash’. The curious chapter included Ron eating Hermione’s family.
In another case, Research Scientist, Janelle Shane shared the text from candy love hearts with a learning algorithm called a neural network to create new messages for sweethearts. See Janelle’s blog post here. Some of the messages generated were fitting for the genre, some were confusing, and others were X-rated.
These examples show that good creative AI generated content may still be a way off. Two things suggest that creative development will remain the domain of humans for the near term. First, great creative content works by generating emotions – something only humans can understand today. Second, the best creative content connects disconnected themes and issues – but only humans can discern whether the connection is nonsense or magic.
A true definition of artificial intelligence sees machines taking on human skills of perception, reasoning, learning and decision-making. Until AI can do these things, marketers should make the most of voice and chatbots, and give humans who will be developing their creative the best possible brief to work from.
Written by Daren Poole, Global Head of Creative Insights Division, Kantar