Can generative AI expand human qualities and connect the world?

Ensuring Ideas Don't Become "Pie in the Sky": Hitachi Driving Social Infrastructure Implementation

Ochi (Dentsu): In Part 1, we discussed the genesis and conceptual approach behind "Today's Whimsical Recipe" (*), the first solution in the "AI for EVERY" initiative that uses generative AI to solve social issues.

Now, let's delve deeper into the "implementation capabilities" of Hitachi and the Dentsu Group. Mr. Ochi from Hitachi, I understand your work primarily involves implementing services and solutions. Could you share how Hitachi typically approaches the process of actually realizing and functionalizing ideas to solve challenges?

※Today's Whimsical Recipe =
Based on Hitachi's inventory management system and supply/demand forecasting and order processing system, it highly accurately predicts "ingredients likely to remain unsold." Generative AI trained on Dentsu creators' insights then generates unique recipes and coupons related to those ingredients. Leveraging the know-how of Dentsu Digital Inc.'s "∞AI Ads," these recipes and coupons are automatically generated as promotional materials and distributed via store apps and in-store digital signage. (Currently undergoing feasibility study)

Ochi (Hitachi): Next, I'll share two examples of Hitachi tackling everyday social challenges. First, Hitachi is developing railway operation systems in collaboration with JR East. This initiative applies AI agents to systems managing conventional line operations and maintenance tasks in the Tokyo metropolitan area, aiming to realize safe and sustainable railway operations.

Hitachi has long been involved in the operation and management of such systems. However, the workforce is shrinking year by year, and system upgrades increase specifications and maintenance complexity. There's also the challenge of ensuring that even less experienced personnel can handle issues promptly when they arise. To address these challenges, we are currently testing the integration of generative AI with manuals and other information, enabling it to assist humans when problems occur. So, you're using generative AI to address both Japan's nationwide workforce shortage and the inherent complexity of the systems themselves.

Hitachi and JR East Japan Agree to Joint Verification for First Use of AI Agents in Railway Operation Management System to Improve Transportation Stability

Second, Hitachi is conducting verification with Daikin on using generative AI to detect anomalies in factory equipment. The AI reads blueprint information to understand the detailed structure of factory equipment. If an anomaly is detected, it checks for issues before and after the anomaly and then provides a likely cause. While still in verification, it demonstrates very high accuracy, performing at a level comparable to that of a typical maintenance engineer.

Daikin and Hitachi Collaborate to Launch Trial Operation Aiming for Practical Application of AI Agent Supporting Factory Equipment Fault Diagnosis

In this way, Hitachi has consistently addressed customer challenges in the B2B domain across different eras by applying diverse technologies. We believe Hitachi's mission is to thoroughly understand the challenges, then combine the technologies we possess—including IT technologies like generative AI, as well as our many strengths in science and technology, such as energy—to work towards solutions.

Ochi (Electronics): Hitachi truly implements services deeply embedded in society, so the dynamism is completely different from the advertising field we primarily focus on! I'd like to ask Mr. Akashi, who has been responsible for design at Hitachi for many years: From the perspective of "design and ideation," how do you perceive Hitachi's implementation capabilities?

Akashi: I've been doing design work at Hitachi my entire career (laughs), so I can't really compare it to other companies. However, since we're a company building social infrastructure, we're extremely careful not to proceed with ideas that are just pie in the sky from the start, with little chance of realization, only to end up achieving nothing. That's the corporate culture I've grown up in. Essentially, if the approach is "the designer came up with an idea, now make it happen," nothing gets realized.

That's why we've always worked together with engineers and architects from the very start, moving forward with a "Ready, set, go!" approach. If we're utilizing AI, we collaborate with implementers like Mr. Ochi.

We also check feasibility early and often. It's not just about technical feasibility; it's equally important to consider potential challenges from a people and organizational perspective when executing. Without jointly planning these preparations, we can't properly judge the merits of the ideas we're currently considering.

Ochi (Japan): So far, we've mainly discussed Hitachi's internal processes. However, for B2B social implementation, it's quite difficult for a single company to achieve this alone. Unless we collaborate with companies actively driving operations in the target field, we risk ending up with just the technology itself.

In other words, I believe a very significant part of Hitachi's social implementation involves working closely with enterprise companies that are actually engaged in operations to build those operations together. It's not just about having the technology.

Dentsu Inc.: Pre-implementation verification using an AI persona with a scale of 100 million people

Ochi (Dentsu Inc.): Next, I'd like to discuss the "implementation" approach within the Dentsu Group. I'm usually involved in creative work using digital and technology at an organization called Dentsu Lab Tokyo... While Dentsu Inc. certainly considers feasibility during ideation, the early stages can be quite innocent, so to speak (laughs). In that regard, I got the impression that Hitachi, which implements social infrastructure, has a much higher awareness of feasibility than Dentsu Inc. Did the two of you from the Dentsu Group notice any differences?

Takahashi: From an implementation perspective, I think the Dentsu Group often sets "moving people's hearts" through communication and expression as the goal. However, how much people's hearts are actually moved relies heavily on tacit knowledge, and it's difficult to promise results because you won't know until you try. Additionally, we often handle communication that lacks similar precedents, making it challenging to guarantee outcomes "after implementation."

However, with recent advances in generative AI, planning that consciously considers implementation aspects is increasing. For example, within the domestic Dentsu Group, we have the " People Model ." This tool creates AI personas representing 100 million people in a virtual space based on large-scale consumer surveys. It allows for simulating ad expressions and conducting mock interviews within this environment.

This approach allows us to simultaneously verify, to some extent, "how likely it is to resonate with people" using generative AI for entirely new expressions, while still launching them as novel concepts into society. While there are still elements where we have to try things out to see what happens, as was the case traditionally, we are increasingly able to test these within the AI environment. This represents the latest movement within the implementation domain at Dentsu Group in Japan.

Kishimoto: Hearing earlier about Hitachi developing an operation system with JR East, I sensed an incredible level of robustness unimaginable in Dentsu Inc.'s creative work.

Speaking about our own implementation, at Dentsu Lab Tokyo, we often plan by aligning feasibility with how to move the hearts of consumers or the people right in front of us, working closely with our "Creative Technologists" – the technical specialists within our creative roles. However, we still can't quite reach the level of infrastructure-level social implementation. In that regard, I think Hitachi's fundamental capabilities are different.

Ochi (Dentsu Inc.): Indeed, I think our respective capabilities and cultures are vastly different. I'm honestly overwhelmed by how Hitachi prioritizes feasibility to this extent, and how their scale, duration, and scope of initiatives differ fundamentally from Dentsu Group's corporate culture.

In this collaboration, we're pinning our hopes on Hitachi's "implementation power." We believe that by working together with Hitachi, we can transform the Dentsu Group's weaknesses—such as short timelines and limited scope for impact— can be transformed into dynamic results by working together with Hitachi.

A future where generative AI expands human potential awaits!

Ochi (Dentsu): Looking ahead, I'd like to discuss how the world might change as we utilize generative AI, incorporating our hopes for the future.

...I wonder how it will unfold. For instance, having AI as store staff, like avatars, could be an entry point. I feel there are various ways AI could play a role. What are your thoughts?

Kishimoto: As the tools keep expanding, I believe AI's capabilities will keep growing. That's precisely why I feel it's important for humans to clearly define their intentions: "What do we want?" and "What would make things better?"

I believe that in the future, using generative AI will enable the creation of more human-centered services, products, and infrastructure. What I mean is, for example, today's so-called multimodal AI can understand multiple inputs like text, voice, and images all at once. It can interpret them on the spot and respond, or to put it simply, it has become quite adept at "reading the room."

As this evolves, AI will become even more adept at empathizing with human feelings, reading the atmosphere of a situation, and expressing itself in more sophisticated ways. Furthermore, as Mr. Takahashi mentioned earlier with the People Model example, AI might eventually be able to simulate in advance how a particular expression will be perceived by people.

As it evolves this way, I think it would be better for AI to aim to "integrate with people" rather than "replace them." The means to achieve that are steadily coming together, right?

Ochi (Denki): I see. This might sound like a strange idea, but perhaps AI could help improve the atmosphere during meetings. For instance, by reading each participant's background and facilitating icebreakers.

Kishimoto: Or if silence falls, it could read the room and jump in, or even comment, "Looks like everyone's getting a bit bored now" (laughs). I think such roles could become possible as multimodal generative AI evolves to capture not just human "words," but also information like "facial expressions" and "sounds."

Ochi (Electrical): What do you two from Hitachi think?

Ochi (Japan): Predicting the future is quite difficult. But what I can say is that the biggest change is probably that interactions with computer programs have shifted from programming languages to natural language.

For example, AI chatbots now grasp the semantic "depth" and "breadth" within natural language quite effectively. This has vastly expanded what's possible, allowing even non-programmers to use AI at a surprisingly high level. I believe this shift is largely due to the interface being natural language.

Moving forward, AI capabilities will continue to expand and evolve. AI agents will communicate with each other and become increasingly capable of solving complex problems. The scope of what can be controlled extends beyond the computer world; for instance, we're now able to simulate human actions like operating a browser with a mouse within generative AI.

Looking even further ahead, communicating with robots might become much, much easier in the future. Today, the world is connected through the internet. Generative AI will likely expand the meaning of "connection" even further. The cost of connecting with the world will decrease, and the depth and semantic richness of information exchange when connected will continue to develop as natural language becomes the interface. What exactly this will become in the future... well, who knows? (laughs).

Ochi (Denki): Hearing you speak, I feel generative AI will enhance its ability to "extend the capabilities of various people." For example, someone with great taste but no programming knowledge could have their ideas translated into designs or applications by generative AI, which could then be purchased by those seeking such solutions... I think this kind of thing will become reality. I feel this will happen across many fields. What do you think, Mr. Akashi?

Akashi: Well, Ochi-san pretty much said everything I wanted to express (laughs). The trend of AI transforming the world as a man-machine interface is absolutely happening, right? The fact that value exchange can occur without manipulating any code, protocols, calibrations, or parameters is incredible. Rather than focusing on the narrative of "AI replacing human jobs and people losing work," I want to really push this discussion forward.

Also, I absolutely love that earlier point about the "cost of connecting people" (laughs). In the first part, I mentioned Hitachi's "Vision Design Project" activities. When it comes to social issues, unless some kind of "new relationship" emerges in society, that value can't be realized, right? Take food loss as a social issue. You can't solve it without creating a "new relationship" between the places where food is wasted and the producers, and then facilitating services through that connection. But until now, the cost of creating that "new relationship" – the cost of "connecting" – was incredibly high.

The gig worker world faces the same issue. For instance, even if I want to exchange work one-on-one with Mr. Takahashi, he doesn't have trustworthy information about me, so he can't trust me. Ultimately, unless someone creates that trust information, gig work or CtoC models won't materialize.

So who connects these dots? It's incredibly difficult to turn that into a viable business. We have the data, right? And we know connecting people improves society, right? But we can't do it. This "frustration of not being able to do it" might become achievable if generative AI lowers the cost of "connecting."

If AI agents become the "connecting piece" for things that were previously too costly to connect within social infrastructure, I think the possibilities would expand tremendously.

Ochi (Denki): That's fascinating! When I think about the efficiency gains generative AI brings, I can't help but imagine, "Won't people just get replaced by AI?" But hearing you now, it really hits home that the true goal of generative AI should be to strongly safeguard the value of human presence, where people connect and match to create new value. That was truly an eye-opening perspective. What do you think, Takahashi-san?

Takahashi: First, as a premise, the trend of generative AI expanding into various areas of life is already underway. And I agree with you that it will become accessible to many more people in the future.

For example, I couldn't write code myself, but now I can ask AI to write Python code for me. By extending human capabilities with AI, all sorts of possibilities emerge. As it evolves further, I imagine a world where, for instance, seniors can simply speak to AI in their own voice and easily access care-related information.

Therefore, I believe the key issue moving forward will be how to distinguish between "what humans fundamentally want to do" and "what can be delegated to AI." Precisely because "AI can do anything," we may be presented with an opportunity to reexamine what humans deliberately choose to do.

In that sense, I hope we can improve society through AI by oscillating between creating systems that safeguard human humanity and, conversely, examining micro-level human perspectives: "What are the fundamental desires humans possess?" and "How will society transform going forward?" Precisely because Hitachi and the Dentsu Group are teaming up, I believe we can achieve the social implementation of AI that stays true to the core principle of "expanding human qualities."

Ochi (Dentsu): Thank you for such a fitting closing remark (laughs)! Hearing this, I truly hope this AI for EVERY project will become an initiative that presents the kind of AI we envision—one that connects people and expands human capabilities and value.

Takahashi: We won't leave people behind, nor will we let it end as a pipe dream. We must tackle this with implementation power. Striking that balance is crucial.

Ochi (Tel): Thank you! Today's conversation was incredibly exciting and truly enjoyable. I look forward to continuing our collaboration.