Why AI is no match for human magic

In a piece first published by Campaign, Newsworks' CEO Vanessa Clifford draws on research into whether AI can accurately predict the winners of the 2018 Newsworks Planning Awards. 

A recipe for advertisers' success

  • Take one brilliant idea based on real human insight
  • Add well-executed creative which speaks to your consumers
  • Mix with relevant contextual placement
  • Serve with meaningful results

When it comes to delivering effective (and award-winning) ad campaigns, this is a basic recipe for success. Not very scientific, I'll grant you, but you don't have to be a genius to work out that the best campaigns include all of these components, executed to the highest possible standard. And, in some cases, the effect is greater than the sum of the parts – a perfect alchemy that comes from idea, creative and context perfectly aligning.

It's these campaigns that grab our attention when we're flicking through the paper, or channel surfing, or traipsing round the West End with armfuls of shopping. Unsurprisingly, it's also these campaigns that often take home the big media awards at the end of a long night.

In my experience, it's generally the campaigns that tap into honest, unfiltered human emotion – whether that be stifled giggles, a creeping tear or stone-cold shock – which stand out, deliver the target results and get the gongs.

With that in mind, is it possible to automate for success? AI is becoming an increasingly dominant feature of our industry, to the extent that it can even be used to script an entire ad. So, if AI can be put to work creating an ad campaign, is it capable of accurately judging a set of industry awards? Is automated learning capable of detecting that special magic produced by the very best campaigns our industry produces? At Newsworks, we decided to put it to the test. 

Unsurprisingly, getting to a definitive answer wasn't an easy feat. When we approached Texture AI with the brief they were both intrigued and anxious. In the words of Dr James Carney, the company's director: "Intrigued, because this is exactly the kind of challenge that draws on our expertise in psycholinguistics and machine learning; anxious because trying to preserve the richness of cultural data when converting it to mathematical form can sometimes feel more like alchemy than science. However, as it turns out, intrigue and anxiety mix better than vodka and tonic, and we learned a lot along the way."

Texture AI analysed five years' worth of Newsworks Planning Award entries in order to create an algorithm for success, which was then put to the test – could it predict the winners of 2018's Planning Awards?

In a word, no. While AI could, with a fairly high degree of accuracy, detect entries which had been shortlisted or highly commended, it struggled to predict the winners that our panel of 20 industry judges had animatedly debated over and ultimately decided on.

As we all know, AI works by establishing patterns and following set commands, so when it comes up against the unexpected it falls short. Yet it's exactly this – the original, unique and novel – which can make the good into something brilliant; creating a spark of magic so rooted in human understanding that AI can't (yet) comprehend it.

So, in the interest of getting some valuable human input, here's what some of our judges thought when we shared the results of the project with them (post-judging day, of course).

Charlie Ebdy, chief strategy officer, Vizeum UK

"Maybe we shouldn't be surprised that the machines would struggle to separate the good from the great. A machine learns patterns, and whilst you would expect good award entries to share certain similarities – an understanding of challenge and solution, a keen grasp of audience behaviour, clearly attributed results – the great often deviate from an established pattern. You want to find the novel solution, work that is surprising, radical or untested, or pure excellence, a level of accomplishment previously unreached. You want something that positively contradicts your expectations.

"What this also suggests, however, is that we could be much smarter in how we use machines to test the strength of our initial thinking and its quality against historic norms. Starting with something the machine thinks might be good gives you a chance to experiment in pursuit of something great."

Paddy Adams, executive director, head of strategy, Manning Gottlieb OMD

"It's an interesting experiment but I'm not sure we'll see awards being handed-out based on the judgment of an algorithm any time soon (other than for the clickbait novelty factor). Much as we like to convince ourselves that advertising is closer to a science than an art and that we're judging entries entirely objectively, the truth is far messier than that. Advertising itself is a subjective skill – it's about people, and people are contrary and unpredictable – and judging awards even more so. Using the criteria of "I wish I'd done that" isn’t exactly scientific, but it's often the best way of working out who deserves to win. On top of that, algorithms can only analyse what has happened in the past and make judgements based on that. I'm not sure how well equipped they are to deal with genuine innovation."

Hai Bei Chen, head of brand, campaigns and reputation, Virgin Holidays

"Whilst the use of machine learning is nothing new for numerical and analytical tasks, I am surprised by the development of it into award judging, and how much the machine agrees with the human judges! However, the outcome confirms that there is no set formula to create that extra sparkle or genuine emotional connection with an audience that the winning entries all did in their own way. This means the machines won't be able to "learn" the secret of winning any time soon. It also means our job continues to be challenging as well as terrific fun!"

What this process shows is that while AI is an increasing force for change within the media industry, it cannot, by its very nature, predict or capture the quirks of innovation. Nor can AI account for the complexities of human understanding and opinion. There is, and will always be, anomalies from the algorithms.

You can take a look at the winners of the 2018 Newsworks Planning Awards here

First published by Campaign.

If AI can be put to work creating an ad campaign, is it capable of accurately judging a set of industry awards?

Vanessa Clifford, CEO, Newsworks
by Vanessa Clifford 03/01/19

Want the latest news, opinion and insight on news brands?
Sign up and we'll send you a weekly update.

Sign up here

Get in touch

3rd Floor
3 Marshalsea Road

0203 949 7470


Keep up to date

Please take a minute to answer a few questions on our new website. The feedback you provide will help us make the website easier for you to navigate and ensure you can find the content you’re after. In addition, we’ll enter you in to a prize draw for a £100 Amazon voucher*.

Take Survey Not now Never