Professor Matsuo, which way will the future of AI and advertising go? (Part 2)

2017: The First Year of Branding Through Digital Advertising

Namikawa: I believe 2017 will be the "Year One of Branding through Digital Advertising." With the ability to deliver rich expressions like video on smartphones, attention is increasingly focused on branding through digital advertising.

I believe we've entered "Phase 2" of digital advertising. "Phase 1" was the data industrial revolution, where companies like Google and Facebook built digital platforms. "Phase 2" is where creativity rides atop that data industrial revolution to move people's hearts. This is a challenging and incredibly fascinating area.

In September 2016, I participated in planning and hosted the event "TCC Kotobamirai Conference." There, I arranged a discussion between copywriter Takashi Nakahata and researchers from IBM Watson (a human-interaction response system powered by learning AI). When Watson reads text, it can identify sad or happy words based on context. Hearing this explanation, Nakahata remarked, "Elementary school boys sometimes tell girls they like 'I hate you'," and while the discussion went nowhere, that moment was incredibly amusing (laugh).

While creativity and technology sometimes clash, humans are a collection of many functions. Among these various functions, there may be some where artificial intelligence outperforms human capabilities.

In Professor Matsuo's book, he writes, "Thinking about artificial intelligence means looking at humanity. It means becoming more self-aware about what it means to be human." I believe that deepening our understanding of humanity will advance our collaboration with artificial intelligence.

Matsuo: Thank you. What you said, Mr. Namikawa, is extremely important. We should steadily implement AI optimization wherever possible. Of course, there are technical challenges, but we can overcome them step by step as we move forward.

Regarding copywriting, it's easy to have multiple drafts created via crowdsourcing, have AI evaluate their quality, retain only the best ones, and evolve the process.

However, there's a long-standing question in AI: "Can a monkey write 'Shakespeare'?" If you let a monkey roll dice from A to Z, eventually it might produce the sequence "Shakespeare." But it would take an enormous amount of time. This illustrates that what's theoretically possible and what's practically usable are two different matters.

Earlier, Mr. Namikawa mentioned the theme of "brand." Since "branding" requires long-term optimization, it might be difficult for AI, which excels at short-term optimization. Humans can build brands quickly using various forms of wisdom, right?

Namikawa: I see. I hadn't considered the perspective that branding with AI is difficult precisely because it requires long-term feedback. Conversely, if short-term feedback is available, even branding could potentially involve greater AI involvement.

Matsuo: For example, I've never climbed Mount Everest. But I can predict it would be so cold I'd likely die if I actually went there. In machine learning terms, this means even without direct experience, I achieve an effect nearly equivalent to having experienced it. That's because I've learned knowledge accumulated over hundreds or thousands of years of human experience through language.

The same applies to advertising domains like branding and copywriting. Humans abstract successful examples from other brands and leverage that knowledge for their own strategies.

Moreover, humans can get inspiration for completely unrelated advertising ideas—like seeing monkeys at a zoo and suddenly envisioning an ad concept that has nothing to do with monkeys. We transfer knowledge across different domains, effectively increasing our pseudo-learning data. Current deep learning technology cannot do this.

Namikawa: So, for now, humans hold an advantage in thinking through long-term feedback and connecting unrelated knowledge. However, brands are ultimately abstract concepts residing in people's minds. Deep learning also involves machines holding abstract concepts during their learning process. In a sense, I think there are similarities.

Matsuo: If we could somehow verify how brand concepts are formed in users' minds, we might be able to efficiently create brands using AI technology.

Namikawa: What would that mean for advertising?

Matsuo: As technology evolves, pursuing overly short-term KPIs will likely lead to deterioration in long-term KPIs. Consequently, technology will probably evolve toward improving short-term KPIs without compromising long-term ones. In that context, I believe we'll gradually see the realization of "moderate comfort and enjoyment."

 

<End>

You can also read the interview here on AdTie!
Planning & Production: Dentsu Live Inc. Creative Unit 2nd Creative Room, Aki Kanahara