From the Front Lines of AI Business: "How Generative AI Will Transform Marketing and Communication" (Part 1)

Amid the emergence of ChatGPT and various other generative AI systems attracting significant attention, the Dentsu Group is developing and providing diverse solutions leveraging AI while exploring the potential of AI utilization from every angle.

This discussion featured Takuya Kodama of Dentsu Group Inc., who oversees the "AI MIRAI" project driving AI adoption across the entire Dentsu Group; Kazuya Kishimoto of Dentsu Inc., also an AI MIRAI member working on AI-driven creative operations; Satoru Yamamoto of Dentsu Digital Inc., engaged in AI-powered marketing support; and Kengo Shibuya of Dentsu International Information Services Inc. (ISID), primarily researching AI×Robotics. These members at the forefront of Dentsu Group Inc.'s AI business discussed industry trends and technologies currently capturing their attention.

An era where we first ask AI about everything, just like we've always searched online

Kodama: First, I'd like to start by hearing how each of you uses generative AI like ChatGPT in your daily work.

Shibuya: Within ISID, a company handling system integrator operations, I routinely use AI to review emails and chat messages, and to scrutinize whether any perspectives are missing during brainstorming sessions. Additionally, as a member of the open innovation organization "Inolab," I'm researching robotics—technology that "moves things." There, I'm also exploring whether we can incorporate large language models (LLMs), the foundation of ChatGPT. Specifically, we're investigating how LLMs could enable robots or apps to linguistically interpret their surroundings and perform appropriate actions accordingly.

Yamamoto: At Dentsu Digital Inc., we developed a solution called " ∞AI (Mugen AI) " that supports advertising creative production with AI. It handles the entire process—from extracting marketing messages to creating creative based on them and predicting effectiveness—using AI. We also utilize AI to streamline development at our Mongolian development company, Dentsu Data Artist Mongol. For instance, we created a chatbot trained on the daily instructions given by management to team members. Using this chatbot enables management-level verification anytime, making AI-driven quality checks during software development now standard practice.

Kodama: It's becoming an era where we ask AI anything or verify things with it, just like how we immediately search online when we don't know something.

Yamamoto: Exactly. Additionally, I use AI to review press releases before publication. It checks whether all necessary elements are included, if there's news value, and if there are any issues from perspectives like fairness, transparency, and ethics towards customers and society.

Kishimoto: In my creative department, it feels like creators have integrated AI into their daily work so seamlessly that we don't even need to mention "we're using AI." Image boards conveying planning concepts, for instance, can now be created astonishingly easily with generative AI. While we still need to refine how AI generates copy ready for direct use in mass advertising, it's invaluable for finding synonyms and rephrasing options.

ChatGPT excels at translation and summarization. When I need to read specialized overseas papers, I often ask it to "explain it in Japanese that even a middle schooler could understand." For abstract or vague questions where search engines don't provide precise answers, it rarely gives direct solutions. However, it's great at providing clues to help you arrive at the answer. What about you, Kodama-san?

Kodama: My primary role is corporate planning, so I frequently read mid-term management plans from companies I want to reference when formulating business strategies. Recently, I had AI summarize various companies' English management strategies, translate them into Japanese, and then extract key points. Using generative AI dramatically simplifies information gathering and improves accuracy. In the future, I expect AI to provide real-time interpretation and translation, create meeting minutes and summaries, making international meetings much easier.

"Alien Intelligence" Generates New Ideas and Value

Kodama: Shibuya-san, don't you think generative AI will significantly boost efficiency in system development workplaces going forward?

Shibuya: That's right. ChatGPT can write programming code, and eventually, even companies without in-house programmers might be able to create their own system templates. However, I don't think the current approach to system development will change drastically overnight. Decision-makers, such as product managers, will still be essential at various stages. But going forward, generative AI could provide numerous hints and assistance for decision-making, potentially accelerating development speed.

Kodama: Nowadays, even without specialized skills, you can create simple games just by providing prompts (instructions given to AI). This suggests prototypes in the early stages of system development could be created much faster than before. Mr. Kishimoto, based on your experience using generative AI in creative fields, do you have any ideas about how AI utilization might evolve in this way?

Kishimoto: I think generative AI will be integrated into various tools like plugins, making the tools themselves more convenient without users consciously thinking they're using AI. Also, besides optimization, AI excels at exploring numerous possibilities based on vast data. As so-called alien intelligence (intelligence distinct from humans), it broadens the scope of ideas, helping humans gain inspiration or utilize it for expression. With the spread of generative AI, I believe this will expand further in the creative field.

Multimodal AI Holds Great Potential

Kodama: By the way, even within generative AI, there are various types—those handling language, images, video, audio, and multimodal systems that can process and correlate multiple data types. Are there any particular technologies or services you're focusing on right now?

Shibuya: I'm interested in multimodal AI, especially technologies that interpret environmental information like images and sounds. Currently, ChatGPT is so advanced that we're experimenting with converting images and audio into textual labels before feeding them to language models. But the information we gather from our surroundings is vast, and much of it can't be easily verbalized. That's why I'm really looking forward to technologies that allow AI to recognize the exact visuals and sounds a person is seeing or hearing, enabling robots to determine what they should do right now. In that sense, I'm also interested in the unreleased feature of GPT-4 (the latest version of OpenAI's natural language processing model) that can directly process images as input.

Yamamoto: Dentsu Inc. has already begun pilot testing chat services that allow interaction with characters. I believe services combining ChatGPT with images and videos will proliferate rapidly. However, regarding audio, serious efforts are just beginning. I anticipate services emerging that create music tailored to the situation within metaverse spaces. As Dentsu Digital Inc., we also aim to develop music-related services within communications utilizing VR and XR.

Kishimoto: Global IT companies are currently researching using ChatGPT to control robotic arms and drones through voice commands, and papers on this have been published. I believe this kind of interaction from text to the physical world holds tremendous potential. As a precursor, there's the theme of how large language models will impact VR, the metaverse, and spaces within games. First, the utilization of multimodal AI within virtual spaces will advance, and then how that will ripple into the physical world. That's where my strong interest lies.

Generative AI like ChatGPT is already commonplace in marketing and creative work. Furthermore, we're now receiving various expectations and inquiries from companies that previously avoided actively adopting AI. In the second part, we'll explore how generative AI will transform marketing and communication, delving deeper into the challenges of scaling it across society.