Trans-N: Developing Systems Free from ChatGPT Lock-In

Trans-N Corporation (Headquarters: Bunkyo-ku, Tokyo; CEO: Harry Na), a company aiming to create interfaces between AI and the physical world, has started offering system development services utilizing open-source large language models (LLMs).

Challenges with ChatGPT

Currently, many businesses are leveraging generative AI across various fields, such as meeting transcription, translation, research, and chatbots for customer support and internal help desks. While integrating generative AI services like ChatGPT API or Copilot has become commonplace, there are several challenges associated with these cloud-based generative AI services:

  1. Information Security Concerns
    Many companies prohibit uploading internal data to the cloud due to information security policies. This restriction often becomes a significant barrier, forcing businesses to abandon the adoption of generative AI services.
  2. Limited Optimization for Specific Domains
    Current cloud-based generative AI services lack the flexibility to adjust their core LLMs. This limitation makes it challenging to deeply tailor AI models to the unique terminologies and needs of internal operations. Tasks that require advanced domain-specific knowledge often cannot achieve optimal performance without the ability to customize AI models to the company’s specific data and needs.

To address these challenges, Trans-N offers systems that ensure data security while maximizing the value of a company’s data assets through the use of open-source LLMs.

Unlocking New Potential with Open-Source LLMs

By the second half of 2024, open-source LLMs have reached or even surpassed GPT-4 in certain performance areas. Leveraging the open-source nature of these models, significant performance improvements can be achieved in specific domains by retraining them on proprietary datasets.

Trans-N proposes a system infrastructure that utilizes open-source LLMs within a private environment. This setup alleviates information security concerns, allowing companies to safely deploy AI. By taking advantage of the open-source nature of LLMs, it is possible to redevelop and retrain the models. Internal knowledge and data collected during AI use are integrated into the AI, enabling it to better understand and evolve with a company’s operations. As a result, a growth cycle is established where the AI provides more accurate responses, increasing user engagement and generating new data.

Trans-N seeks to collaborate with companies that share an interest in and support for the vision of utilizing open-source LLMs.

Building Future AI Agent Systems

Expanding upon the above methods, Trans-N is working on developing various agent systems for businesses. If we consider AI as a "brain," we aim to equip it with "eyes," "ears," and even "hands" and "feet," enabling AI to operate effectively in the physical world. By connecting numerous AI agents to an AI “brain,” we enhance human knowledge, decision-making, and support real-world actions.

Trans-N’s agent systems promise to continually acquire advanced and general AI capabilities, providing optimal support for a variety of physical environments. Our vision of creating an "interface between AI and the physical world" will transform work processes, enabling people to focus on more creative activities.

We are already collaborating with a leading Japanese general trading company on these agent systems. Stay tuned as we gradually unveil these initiatives. Let’s build a future where AI comprehends, connects with, and utilizes the physical world.

About Trans-N

Trans-N, founded in 2024 and headquartered in Tokyo, is an innovative company dedicated to driving digital transformation for Japanese businesses through AI technology. Established by a team of experts with international backgrounds, the company integrates cutting-edge AI technologies and best practices from around the world into the Japanese market, delivering tailored AI solutions across industries.