Panasonic Connect and DeepL: leveraging Language AI for better global communication

What you need to know about Panasonic Connect:

• Established in 2022, Panasonic Connect Co. is part of the Panasonic Group, a Japanese multinational electronics company based in Osaka, Japan
• Panasonic Connect plays a pivotal role in the growth of the company’s B2B solutions business
• Panasonic Connect uses DeepL for both translation and writing improvement, allowing its teams to communicate with confidence across borders

Recently, Tak Shirai, Country Manager for Japan at DeepL, sat down with Shoji Otsubo, Senior Manager and AI Technology Evangelist of the R&D division at Panasonic. They discussed how Panasonic Connect uses DeepL’s Language AI technology to break down language barriers within their organization.

Let’s dive in.

Mr. Otsubo, could you please explain your current role in the company and the scope of your responsibilities?

I’m a member of the R&D department at Panasonic Connect. As a Senior Manager, I’m responsible for the research and development of AI for use in a wide variety of fields and enterprises. I’m also working as an evangelist who communicates our technology and AI initiatives to other customers.

Panasonic Connect Corporation is a member of the Panasonic Group. Originally an in-house company within Panasonic Corporation that was in charge of B2B solutions, it became independent as a legal entity when the company was reorganized into a holding company. Our customers are corporations and government agencies in particular.

Our mission is to create change in society from the bottom up and to connect this change to future development. Originally, we started our business with hardware edge devices, and we have worked directly with a wide range of customers. However, our customers were not looking for hardware—they were looking for solutions. 

For customers in the supply chain, public services, lifestyle infrastructure, and entertainment sectors, we originally provided technology for devices such as projectors and software. Now we also provide consulting services to understand the customer's business processes.

I’m in charge of the Technology R&D division, meaning I’m engaged in research and development focusing on the following technological areas: 

• The digital domain: such as the so-called "inside of PCs and the cloud"
• Sensing: which is what humans see and hear in this physical world in which we live
• AI and simulation: for sensor research/exploration
• Robotics: which offers the technologies we develop as a service to customers who really need them

We’re also working on a scheme to expand our services. For example, Blue Yonder, a U.S. company we acquired in 2021, provides infrastructure system software services to supply chain companies. We’re extending the service by incorporating the data we collect in the field.

Could you discuss the challenges of breaking language barriers?

Panasonic Connect has a global customer base, and with the acquisition of Blue Yonder, as I mentioned earlier, we now have an even greater number of overseas customers.

Also, our CEO, Yasuyuki Higuchi, encourages promoting management and internal administrative reforms. We also announced the release of our own generative AI, ConnectAI. Additionally, we’re collaborating with various external experts to change the way we do research and development itself.

My boss has been in Germany since last year, and I’m constantly communicating with him to discuss the next technological and AI strategies. In addition, more and more of my fellow researchers are based globally, including members from Africa and China.

In this environment, communication issues naturally arise on a daily basis. Especially in terms of management of the R&D department, how we keep researchers motivated is important. Research is a field where it’s not easy to get things right: you may or may not get one hit out of every ten.

In this situation, it’s necessary to properly communicate the intent of instructions from the supervisor to the members of the team. If we don’t communicate appropriately when changing research policies, for example, a single word can cause an unforeseen outcome due to a major difference in perception.

In addition, the research area of AI in particular evolves rapidly. We live in a world where technology from one week ago is obsolete one week later—and a new technology has been developed. I have to read many papers every day and present them at conferences. In this situation, reading and writing in English slowed me down significantly.

With such language challenges, I assume you are using the translation function of DeepL Pro. Can you tell us how it has changed your business?

Using DeepL has really changed our world. There are two major benefits. Firstly, as those of you who have used the service may have realized, the translation is as natural as if you were really writing in Japanese. There's almost no awkward translation from English to Japanese, or vice versa—as is often the case with conventional free services. 

Second is the speed of translation. It is, of course, fast, but there’s also the fact that various user interface innovations have been made. For example, it’s now possible to communicate with different languages in a breathtaking way—by simply typing the shortcut "Control CC" and the selected text is immediately translated.

I use a browser wiki to communicate with my boss to prevent miscommunication. I write sentences in Japanese, but I have DeepL in my browser extension, so sentences written in Japanese can be translated directly into English. My boss also has an account on DeepL, so I don't know if he sees it in English or German, but he can also see the Teams browser screen and the Wiki screen in his preferred language.

It’s a big change to be able to communicate clearly and quickly in multiple languages—it’s truly breathtaking.

Recently, DeepL Write Pro was announced in Japan ahead of the rest of the world, and Panasonic Connect were using the beta version of DeepL Write Pro. Can you tell us how you use it?

I use English most when I am writing research papers. I’ve been evaluating DeepL Write Pro with members who also write papers, and the first thing they say is that it’s actually useful.

I asked them how they usually write papers in English, and many of them would write something in Japanese and then translate it into English. The common process was to ask a paid human editing service to improve the English text and submit it to an international journal. 

However, when we used DeepL Write to edit a paper that had already been proofread by the paid editing service, we found that DeepL Write provided 5–6 times more editing suggestions than the paid service, which made 2–3 corrections per page. I find the alternative suggestions to improve the text from DeepL very effective.

Another member of my team used a generative AI service to translate and edit the papers he writes for overseas journals. When I evaluated DeepL Write with him, he told me that the problem with using a (general) generative AI is that, if you change the nuance slightly, it changes the text to something completely different from the previous translation. 

This means you lose track of the changes and correspondence between the modified parts, making it difficult to control. When we edit using DeepL Write, we can clearly see which section corresponds to which section, and we can correct as intended. For us, this is one of the most powerful aspects of DeepL Write.

Could you share what kind of world you think exists beyond language barriers?

Our research and development includes generative AI, and we’re very interested in the technological trends that we’re investigating at the moment. I think further evolution in AI will take place in the near future. 

One possible evolution is that we’ll be able to make logical inferences, which I think will happen before the end of March 2025. Not just any results, but results supported with increased logic. 

Secondly, I think that personalization is a possibility. AI can respond to what each person wants to say and the nuances they want to convey, and at the same time, the data will be securely protected.‌ I believe the balance between personalization and security will probably be achieved as a result of technological evolution.

In addition, while multimodal AI is also emerging, its understanding of context will be improved. I believe that in the next few years, we’ll see an evolution in the way that the context of a translation is interpreted and then expressed appropriately.

However, we will only truly communicate across diverse communities when people of different languages and cultures can understand the meaning of implicit or culturally nuanced expressions. These expressions, which are implicit in our minds and concepts, differ by culture. 

I believe when this is achieved, it will be a true transcendence of diversity and direct communication. Beyond the language barrier, values and culture will be the next step, and I’m really looking forward to this. I’m sure this will be technically very difficult, but I’m very much looking forward to seeing it progress to that point.

Want to talk to our experts about how your company can use Language AI to communicate globally? Our Sales team would be happy to discuss.