Ichiro Jinnai, Director of Out of Home Media Services Division at Dentsu Inc., hosted a panel on Artificial Intelligence in the real world at Advertising Week Asia, with a focus on the ways AI can be monetized. Japan has 10% of its population in the cities, and this is estimated to grow to 50% by 2020. Current metrics and usage of IoT predicts that AI will help make the cities more efficient, able to handle the population growth, and even AI will be able to predict future trends.
“AI has an indispensable relationship with IoT” —Motoaki Nishiwaki, Evangelist (Executive Officer), Microsoft Co.
Microsoft worked on a project with London Underground subway based on using AI to predict malfunctions. For example, after accumulating data on an escalator’s mechanical information, such as oil, hours of usage, age of parts, etc., and comparing against social data of users in the station’s area of that escalator, it could be predicted when the equipment would fail. Microsoft’s studies also found that the AI predictions were more accurate than relying on equipment engineer’s predictions. Microsoft has not only done these studies on escalators in the London Underground, but also on elevators and vehicles in the United States. “There is a huge amount of things connected to the IoT and collecting a ton of data, and by using that data, AI is predicting the future.”
Hiroshi Ohta, President & Co-Founder of Cloudian presented a more specific and in-use AI campaign.
The Deepad Project test case took place in Roppongi area. The large billboard up on the building used an attached camera to take photos of the cars in traffic. The AI would identify the manufacturer, model and year of each car approaching, and the billboard would change the displayed advertisement to match the profile of the car model owner.
The AI would show a golfing ad on the billboard for a luxury car, but for a family-model car the ad would change to something more appropriate. The image above shows the AI process: “‘Deep Learning’ is AI requiring training.” For example, the AI platform has 337 models with over 5000 per model, and the AI is brought close to the location where the camera will go, like a parking lot, and the AI will train on the cars in the parking lot. For Cloudian, OOH advertisement application of AI is already in use, and collecting traffic as marketing data is soon to start.
“Responsive Facial Recognition uses deep learning algorithms to anonymously detect gender, facial expression, age, and composition of passing audience to serve responsive content on an advertising screen,” Ben Milne, Managing Director, Posterscope Japan, explains. Posterscope uses this platform to make OOH digital “AI Poster” advertisements that can “read” the audience and adjust the display in real time. Posterscope then gave the AI algorithms to adjust the “DNA” of the poster and experiment with images and text. As the AI learned and gathered data in regard to what was most effective in the passing, public audience, Posterscope found not only the expected peaks and valleys in audience response, but that the AI found some images to be effective that perhaps normally an agency wouldn’t think to use. For example, an image of a dolphin was effective in a coffee campaign, but it’s unlikely that a person at an agency would think to use a dolphin in a coffee advertising campaign. Ben poses the question, what will it take, or how will it come to be, that a person could trust an AI to make correct creative decisions?
It is expected there will be a paradigm shift in the marketing and advertising industry in the usage of big data and how it feeds AI, but also the monetization of using this technology for targeted advertising will change OOH advertising.