PEARL

TitlePEARL
BrandBELGIAN ASSOCIATION OF MARKETING
Product/ServiceMIXX AWARDS
Category A04. Early Stage Technology
Entrant DDB BRUSSELS, BELGIUM
Idea Creation DDB BRUSSELS, BELGIUM
PR DDB BRUSSELS, BELGIUM
Production DDB BRUSSELS, BELGIUM
Credits
Name Company Position
Peter Ampe DDB Brussels Creative Director
Odin Saille DDB Brussels Creative Director
Francis Lippens DDB Brussels Business Director
Sara West DDB Brussels Concept provider
Silke Beurms DDB Brussels Copywriter
Stefanie Warreyn DDB Brussels Head of digital production
Maria-Laura Laubenthal DDB Brussels Producer
Kenn Van Lijsebeth DDB Brussels Copywriter
Jorian Vanvossel DDB Brussels Strategic Planner
Sven Verfaille DDB Brussels Designer
Jos Polfliet Faction XYZ Head of Machine Learning
Joeri Van Steen Faction XYZ Chief Technology Officer
Milan Gada Microsoft Program Manager Cognitive Services
Maarten Baert Dividebyfour 3D director
Rik De Lausnay Dividebyfour Lead 3D artist
Gabriel Hennessy Dividebyfour Houdini artist
Peter Baert Raygun Managing Partner - Composer
Lucy Akhurst Raygun Actress

The Campaign

Could artificial intelligence make awards judging fairer in the future? Introducing Pearl, the first AI jury experiment. This first AI experiment on judging will investigate if creativity is measurable and how AI might be able to help the judging system. Named after Radia Perlman, one of the prominent female internet pioneers, the result is Pearl. Pearl was fed with all international MIXX case studies from the last decade and given 3 months to analyse the texts, videos, results and even music, making her surely the most experienced advertising jury on the planet. Pearl uses state-of-the-art technology to learn from all of those cases and make her own prediction as jury member of the award show.

Creative Execution

Together with Faction XYZ, a Belgian shop specialized in machine learning, we gathered over 30.000 entries from the last decade of MIXX awards and their respective results. This allowed us to feed Pearl with all the relevant data to learn what works and what doesn’t. Using Amazon’s High Performance Computing solutions in combination with an early version of Microsoft’s Video Indexer and Spotify’s API, she could distil elements such as voices, subject, music, dialogues and even originality. These quantified elements in combination with the known results of all those 30.000 entries allowed her to learn about what makes or breaks a case. At this point in time, Pearl has been used to judge all of this year’s MIXX entries as a rest run. In the end she chose the same winner as the human jury (Pearl has a 76% accuracy when predicting cases). As of now Pearl is available to be used at other award shows as well.

There still is room for improvement as she needs to be able to better recognize humour for example. But Pearl has already proven helpful in e.g. detecting original combinations to look for creativity. Pearl is 76% accurate: Out of all MIXX Award entries, Pearl picked the same winner as the human jury did. Proving that AI can be helpful as an additional tool for award show judges: she never grows tired, never loses attention and doesn’t have friends to defend or a country or network to promote. Reaching over 14 million advertising professionals, Pearl raised a debate about the future of creative judging:In larger competitions she can be helpful in pre-judging to avoid that good campaigns get overlooked. She isn’t there to replace human judgement, but she can keep an extra eye out for them. In the future, we hope to let Pearl assist human judges in other award shows.

Pearl is an AI jury experiment which distilled and quantified data from over 30.000 award show entries. Based on the results of all those entries, she created an algorithm which allowed her to predict award show success with a 76% chance of success. To do so successfully she was granted early access to Microsoft’s Video Indexing software.