the first Japanese artificial intelligence dialogue system, capable of speaking and listening simultaneously, is now available to the public

Publié le 16 July 2025 à 09h29
modifié le 16 July 2025 à 09h29

The first dialogue system in artificial intelligence from Japan revolutionizes human interactions. Designing a machine capable of listening and speaking simultaneously opens new technological perspectives. Through this innovation, natural conversations reach new heights.

An authentic linguistic exchange model is taking shape, propelling interaction between machines and humans towards unprecedented mutual understanding. The stakes related to the integration of this technology multiply across various fields, from education to customer assistance.

This technological advancement promises to transform daily communication by assimilating the essential cultural subtleties of the Japanese language.

A turning point for Japanese artificial intelligence

The development of J-Moshi, the first dialogue system in artificial intelligence from Japan, marks a significant advance in the field. This system is designed to faithfully reproduce Japanese conversation patterns, which include dynamic exchanges through short replies known as aizuchi. These interjections, such as Sou desu ne (that’s right) and Naruhodo (I see), enrich verbal interactions.

Design and development of J-Moshi

J-Moshi was developed by the Higashinaka Lab team at Nagoya University. By using the Moshi model in English, the team made good use of a comprehensive dataset. The training process lasted about four months, integrating data from J-CHAT, the largest Japanese dialogue dataset, collected by the University of Tokyo.

This ambitious project required the use of 67,000 hours of audio from various media such as podcasts and YouTube videos. These results are the product of rigorous collaboration between academic research and market needs.

Practical applications

J-Moshi has generated significant interest, particularly for its potential in language learning. Its interface provides non-native speakers with an exceptional opportunity to practice the nuances of Japanese conversation. Its use is also emerging in sectors such as call centers, the medical field, and customer service.

The challenges related to adapting J-Moshi to the specifics of these fields remain. The prevalence of English data complicates the implementation of specialized systems, making resources for learning the Japanese language more limited.

Challenges and technological innovations

Professor Ryuichiro Higashinaka, head of the laboratory, highlights the unprecedented challenges encountered in Japanese AI research. Concerns regarding privacy and the shortage of voice resources present significant obstacles. To address this, the laboratory has developed innovative solutions, such as using AI programs to isolate voices in audio recordings.

Dialogue systems currently hinder complex interactions, particularly the interpersonal relationships that come into play in various contexts. Visual aids like masks can distort perceptions by concealing essential cues such as facial expressions. Field tests at the NIFREL Aquarium in Osaka have demonstrated the need for human intervention to address complex issues.

Future perspectives

The development of J-Moshi opens doors to harmonious collaboration between humans and machines. Guide robots, like those deployed at the Osaka Aquarium, are capable of handling routine interactions while being able to quickly switch to a human operator if needed. This approach is integrated within the national Moonshot project, aimed at improving service quality through advanced AI systems.

Researchers continue to refine human support systems. Conversational analysis methods and dialogue disruption detection systems have emerged, facilitating quick responses from operators during system failures. Current research extends beyond J-Moshi, encompassing various methods for interaction between humans and robots.

Recognition and publications

The success of J-Moshi in the field of artificial intelligence has been confirmed by its acceptance for publication at Interspeech, one of the most influential international conferences dedicated to speech technology. The work of Professor Higashinaka and his team will be presented in Rotterdam, The Netherlands, in August 2025.

This advancement in dialogue processing highlights the future potential of AI systems. The development of technologies enabling fluid interaction between humans and machines is now within reach. This paves the way for major innovations, positioning Japan as a central player in the development of dialogue systems in artificial intelligence.

Frequently asked questions about the first dialogue system in artificial intelligence from Japan

What is J-Moshi?
J-Moshi is the first dialogue system in artificial intelligence in Japan, designed to reproduce natural conversation patterns in Japanese, capable of talking and listening simultaneously.

How does the J-Moshi system work?
J-Moshi uses dialogue data and Japanese voice datasets to train its model, allowing for natural interaction by integrating characteristic interjections known as aizuchi.

What are the advantages of J-Moshi compared to other AI systems?
Unlike traditional AI systems, J-Moshi manages contextual responses and adapts its interactions in real-time, thereby improving the user experience for Japanese speakers.

How is J-Moshi used in public environments?
The system is used in places such as the NIFREL Aquarium in Osaka, where it provides answers to visitors while allowing human operators to intervene for more complex questions.

Can J-Moshi be used to learn Japanese?
Yes, J-Moshi can help language learners practice natural conversation patterns, making the learning process more interactive and dynamic.

What challenges does J-Moshi face in its development?
Challenges include the shortage of data for Japanese voice and complex social situations that can hinder the system’s understanding and management of interactions.

Are there any commercial applications planned for J-Moshi?
Yes, the research team is exploring applications for call centers, health services, and customer service, although adapting to specialized fields is complex.

When was J-Moshi launched to the public?
J-Moshi gained popularity in January 2024, when demonstration videos went viral on social media.

How will the technology evolve in the future?
Professor Ryuichiro Higashinaka anticipates that systems like J-Moshi will be able to work in fluid collaboration with humans, integrating gestures and natural communication.

actu.iaNon classéthe first Japanese artificial intelligence dialogue system, capable of speaking and listening...

translated_content> Taco Bell reconsiders its choice to use AI at the drive-thru

découvrez comment taco bell reconsidère l'utilisation de l'intelligence artificielle pour améliorer l'expérience au drive-in, entre innovation technologique et attentes des clients.

A smart electrocardiogram: the revolutionary tool to help doctors detect heart diseases in record time

découvrez comment un électrocardiogramme intelligent révolutionne le diagnostic cardiaque. grâce à cette technologie innovante, les médecins détectent les maladies cardiaques plus rapidement et avec une précision accrue, offrant ainsi des soins optimisés aux patients.

Taco Bell slows down the rollout of its smart drive-thrus after technical issues and inappropriate behavior

taco bell ralentit le déploiement de ses drive-in intelligents en raison de problèmes techniques et de comportements inappropriés, mettant ainsi en pause sa transformation numérique pour garantir la sécurité et la qualité du service.

Towards a new era of artificial intelligence: the emergence of interactive agents in customer relations

découvrez comment l'intelligence artificielle transforme les relations clients grâce à l'émergence des agents interactifs : des solutions innovantes pour personnaliser et améliorer l'expérience client.

Google Meet: Guide to activate the voice translation feature in French

découvrez comment activer facilement la traduction vocale en français sur google meet grâce à notre guide complet. simplifiez vos réunions multilingues en quelques étapes simples !

Artificial Intelligence: A Tool, Not an Escape for Writing

découvrez pourquoi l'intelligence artificielle doit être considérée comme un outil d'aide à l'écriture plutôt qu'une échappatoire, et comment elle peut enrichir votre créativité sans remplacer l'essence humaine.