Winning First Place in an International AI Competition: Top Engineers Discuss the Future of Dentsu Group × AI (Part 2)

• Winning First Place in an International AI Competition: Top Engineers Discuss the Future of Dentsu Group × AI (Part 1)

In May 2022, AI engineers Ryuichi Ishikawa and Hideki Murata secured first place among 1,471 teams in the "NBME - Score Clinical Patient Notes" AI competition on the global AI competition platform "Kaggle."

The Advanced Creative Research Center (ACRC) at Dentsu Digital Inc., where both are affiliated, is an organization within the Dentsu Group Inc.—often perceived primarily as an advertising agency—that aims to fuse data/AI (artificial intelligence) with creativity.

This is the second part of our interview with the two, exploring how they leverage their world-class AI engineering skills in their work at Dentsu Digital Inc. We discussed AI utilization unique to the Dentsu Group, the potential of AI × Creativity, and the first steps companies should take to begin utilizing AI.

The key lies in how to make digital advertising—which has limited expressive range—more creative.

Q. What areas would you both like to strengthen further going forward?

Ishikawa: Compared to mass advertising, digital advertising is often said to have a narrower range of expression. Personally, I also feel digital advertising can lack appeal in some areas. That's why I want to use AI to make digital advertising more creative.

For example, mass advertising has aimed to move consumers with a single piece of copy. Digital advertising, however, creates 100 pieces of copy that are 60 out of 100 points. The idea is that the inefficient ones get weeded out, leaving only the efficient ones. If you look at just this part, mass advertising, which competes with one top-notch piece of copy, could be considered more creative.

But as digital advertising grows this dominant, the approach shifts toward delivering ads optimized for each individual, not just the masses. Sometimes, appealing with "Free shipping right now!" might lead to a purchase from someone seeing a banner ad, more effectively than a universal message. The nature of advertising itself is shifting. We must engage with creativity while acknowledging this reality.

Murata: I'm involved in predicting the performance of banner ads using AI, but we're still only halfway there. We've only just begun automating banner generation with AI, so I want to further refine it to make as many companies as possible want to use it.

Considering AI applications unique to Dentsu Inc. within the creative context

Q. Mr. Ishikawa, you seem intent on linking AI with creativity to generate interesting expressions. However, AI excels at prediction and statistics, and it's also a technology that drives efficiency. If efficiency advances, doesn't that move us further away from interesting content? What do you see as the unique creativity that AI can offer?

Ishikawa: It's true that as AI becomes more optimized, there's an aspect where it ultimately leads to the same destination. For example, when you follow the recommendations of a YouTube engine and keep watching videos, don't you often find yourself ending up watching the same types of videos? Personally, I sometimes feel my perspective is narrowing because of AI.

That's why I'm now proposing an AI that deliberately introduces you to things far removed from your usual choices. Haven't you ever picked up a DVD that caught your eye at a rental shop and unexpectedly discovered a great film? It would be fascinating if AI could recreate that kind of serendipity. Rather than just optimizing recommendation engines, incorporating creative elements could deliver unprecedented experiences. I believe this fusion of AI and creativity represents the unique expression possible within the Dentsu Group.

Murata: Conversely, I wonder if AI itself needs to be creative. We have many planners at our company. If AI streamlines their tasks, they can dedicate that freed-up time to creative work. That's also a form of creative AI utilization, right? We aim to pursue that direction as well.

Q. This overlaps with what Mr. Ishikawa just mentioned. If we pursue AI solutions to their logical conclusion, we might all arrive at the same destination. In that context, what direction do you think would differentiate us and allow the Dentsu Group to offer truly distinctive AI solutions?

Ishikawa: I believe leveraging AI within a creative context is what defines the Dentsu Group. For example, " TUNA SCOPE," which won a Bronze Award in the Product Innovation category at the international advertising awards "Clio Awards," is very Dentsu-like. I was thrilled by the concept of an AI that inherits the skills of expert craftsmen through deep learning to assess tuna quality. While AI for operational efficiency is necessary, as you pointed out, differentiating solely by performance is difficult. I believe the Dentsu Group's distinctiveness lies in the growing number of other unique AI applications.

Leveraging existing systems enables low-cost AI implementation

Q. I imagine many companies feel, "We want to try AI, but we don't know how." Could you share an effective approach for such companies from an AI engineer's perspective?

Ishikawa: Sometimes relying on technology yields better results than relying on human intuition. That said, AI isn't a panacea. If expectations are too high—thinking "AI can do anything"—then even one incorrect prediction or analysis might lead to the conclusion "AI doesn't work," potentially causing companies to shy away from using it in the future. Therefore, I believe it's best for them to understand that "it's accurate about 7 out of 10 times" or "it's more precise than humans," and then continue using AI consistently.

Q. Many clients without AI implementation experience likely have no idea about schedules or costs. Wouldn't knowing these metrics make it easier for clients to consider adoption? For example, could you provide a rough estimate like, "Using existing systems, you can implement AI within this budget range"?

Ishikawa: Many people assume AI solutions are costly. Developing a completely new system from scratch does indeed take three to six months and involves significant expense. However, the Dentsu Group offers numerous AI solutions, some of which can be implemented starting at several hundred thousand yen.

For instance, tools like " ADVANCED CREATIVE MAKER®︎ "—which uses AI for automated banner ad generation, effectiveness prediction, and analysis—and " TREND SENSOR" —a system that analyzes social media and mass media data with AI to predict upcoming trending keywords—can be utilized without significant cost. We hope that through these systems, people will first experience "Oh, AI can do this?" and "AI is interesting!" and that a culture of continuously using technology will emerge.

When it comes to AI adoption, businesses likely expect "improved operational efficiency" most of all. At Dentsu Digital Inc., we aim to add uniqueness by combining this with creativity, further expanding AI's potential. Approaching AI with unique ideas, like recommendation engines incorporating randomness, might reveal AI applications and competitive advantages unseen when viewed solely through the lens of operational efficiency. We hope companies hesitant about AI adoption will start by utilizing existing AI solutions to experience its potential.