Samsung Electronics hosted its annual Samsung Developer Conference Korea 2024 (SDC24 Korea), a virtual event showcasing cutting-edge advancements in software, artificial intelligence, IoT, and open-source development. At the event, Samsung unveiled its next-generation AI model, Gauss2, highlighting its enhanced efficiency and versatility for various applications.
Samsung Gauss2 builds on its predecessor, the company’s first in-house AI model, with improvements in handling multimodal data such as text, code, and images. Gauss2 comes in three versions tailored to different use cases:
• Compact Model: Optimised for devices with limited computing power, offering on-device AI functionality. • Balanced Model: Strikes a balance between speed, performance, and efficiency, making it suitable for diverse tasks. • Supreme Model: Utilises advanced “mixture of experts” (MoE) technology to lower computational costs without compromising performance.
“Gauss2 is already boosting our internal productivity, and we plan to integrate it into products to deliver higher levels of convenience and personalisation,” said Paul Kyungwhoon Cheun, President and CTO of Samsung’s Device eXperience (DX) Division.
The Gauss2 model supports between 9 and 14 languages, along with multiple programming languages. Samsung has introduced custom stabilisation techniques and a tokeniser to enhance its efficiency, especially in handling complex tasks like coding and multilingual translations.
Samsung has already incorporated Gauss2 into its operations, empowering internal tools like code.i, an AI coding assistant for the company’s developers. Since its initial release last December, the tool has seen a fourfold increase in monthly usage, with approximately 60% of Samsung’s Korean DX Division software developers now actively leveraging it.
Additionally, the Samsung Gauss Portal, a conversational AI assistant, is helping employees streamline tasks such as document summarisation, translations, and email composition. The service has been rolled out to international subsidiaries, while customer service call centres now use Gauss-powered tools to categorise and summarise customer interactions automatically.
Samsung plans to expand Gauss2’s capabilities, including enhancing its natural language processing, multimodal functions like interpreting charts and tables, and supporting image creation. Under its “AI for All” vision, Samsung aims to integrate Gauss2-based services into its entire product lineup, enabling users to enjoy smarter, more personalised interactions with their devices.
The company is also working to combine AI with knowledge graph technology, promising even deeper levels of customisation across its devices and ecosystems.