In this final installment of Samsung’s AI Leadership series, Samsung Newsroom explores how Samsung’s technological expertise and research initiatives aim to refine AI-driven user interactions across its expansive portfolio.

Prioritizing Customer Experiences

Samsung utilizes on-device and cloud-based AI solutions in its offerings. This strategy is further reinforced by the pursuit of hybrid AI, which integrates the strengths of both technologies to provide context-aware, tailored solutions for diverse user needs. The decision on which type or combination of technologies to employ is anchored by this ultimate goal — “How can we best serve the consumer?”

On-device AI delivers secure and efficient AI experiences in devices such as smartphones and TVs by utilizing only the data and resources on the device itself. Because on-device AI doesn’t communicate with the cloud, responses are virtually instantaneous. Since data doesn’t leave the device, the technology is often considered safer for sensitive procedures such as those that involve personal information.

Similarly, edge AI operates independently of cloud infrastructure and relies solely on specific local resources to perform AI functions. In interconnected home environments where appliances and IoT devices work in tandem, data processing in edge AI can be handled by either devices with higher performance specifications or devices nearby — without depending on the cloud. Thus, edge AI offers speed and security benefits akin to on-device AI.

While on-device and edge AI rely on technologies that use smaller models and fewer computations in constrained hardware environments, experiences that require access to extensive online information or high-performance computing are supported by cloud-based AI. The technology utilizes vast datasets and large-scale AI models hosted on servers to deliver a range of advanced capabilities. To achieve high performance while maintaining efficiency and security, Samsung conducts extensive research in areas including minimizing data latency, optimizing model efficiency and implementing data anonymization and encryption.

With expertise in all these different forms of AI — on-device, edge and cloud-based — Samsung takes a step further and combines different formats as necessary to best serve the needs of each use case. In hybrid AI, sensitive and quick-response data can be processed using on-device AI models while the cloud can manage processing that requires external resources like up-to-date information or high-performance models. Galaxy AI, for instance, uses both on-device AI and cloud AI — operating either independently or simultaneously as needed — to deliver optimized solutions for users.

▲ Daehyun Kim, Head of Global AI Center, Samsung Research explains Samsung’s AI Strategy.

Generative AI Integrated Into Everyday Life

Samsung continues to drive innovation in the realm of generative AI, further enhancing the versatility of its products and services. For instance, Galaxy AI offers a suite of features¹ including Chat Assist, which supports translation and interpretation in select messaging apps; Note Assist, which helps summarize notes and generates note covers; and Photo Assist, which edits images with generative AI.

Generative AI is also supported on select Samsung TVs and home appliances through the Generative Wallpaper feature.² Users can enjoy 4K wallpapers that fit their tastes on the 2024 Neo QLED TV or transform their kitchens with AI-generated cover screens on the Family Hub refrigerator.

▲ Generative Wallpaper on Samsung’s AI TV

To offer the best generative AI experience on its devices, Samsung not only collaborates with partners but also develops its own AI models. From fundamental technologies beginning with initial research — including foundation model development — to data collection and training as well as lightweight AI model development, Samsung leverages its vast expertise across the AI R&D board to optimize solutions and meet each product’s specific needs.

For example, Samsung developed proprietary foundation models specifically for image generation, editing and transformation to support a wide range of visual creation experiences. This was a challenge because unlike text data — which is composed of discrete values — image data consists of continuous values, offering nearly limitless possibilities for generation. To address this complexity, the company applied specialized training techniques to the image models while reducing their size for efficient on-device operation. Developers refined deep learning algorithms and data augmentation techniques to allow for rapid computation and seamless image creation with minimal processing requirements.

These foundation models were then fine-tuned for specific applications. For instance, the Photo Ambient wallpaper in Samsung’s latest smartphones integrates the image model with real-time data to create dynamic wallpapers that visually reflect current conditions such as snowfall or a starlit night depending on the time and weather.

Samsung Gauss2: Samsung’s Proprietary Generative AI

The application of generative AI is rapidly expanding beyond personal daily use into enterprise environments. Samsung has developed a proprietary generative AI model to strengthen its competitiveness in AI technology, securely manage sensitive internal information and optimize performance and scalability for specific purposes. Proprietary generative AI allows the company to establish models tailored to a wide range of products and services while improving internal productivity in a more secure environment.

Samsung recently unveiled its upgraded proprietary generative AI model, Samsung Gauss2 with better performance and efficiency compared to its predecessor, Samsung Gauss1, introduced last year. Samsung Gauss2 is a multimodal³ model capable of simultaneously processing and understanding various data types.

Samsung Gauss2 is available in three models tailored to specific use cases — Compact, Balanced and Supreme. The Compact model is a lightweight variant optimized for on-device AI. The Balanced model, a medium-sized cloud-based variant, is designed for tasks such as text and code generation. The Supreme model, a large cloud-based variant, is built to support high-performance services with a particular focus on improving workplace productivity.

▲ Samsung unveiled its upgraded proprietary generative AI model, Samsung Gauss2.

To efficiently implement large models like the Supreme variant, Samsung has incorporated the Mixture of Experts (MoE) technique to selectively activate only the “expert models” most suitable for specific tasks within a large-scale neural network.

MoE significantly reduces the number of computations during training and inference by activating only the necessary components, rather than running all expert models simultaneously in response to each query. This selective activation not only enhances efficiency but also facilitates the expansion of the model to address increasingly complex problems by adding more expert models. Samsung has meticulously trained these expert models to operate accurately and without interference.

▲ MoE optimizes computation to implement large AI models more efficiently.

Samsung Gauss2 is already being applied across multiple areas within Samsung. For instance, the Device eXperience (DX) Division employs Samsung Gauss2’s coding assistant service to support research and development activities. This service leverages a specialized model — trained on Samsung’s proprietary data, code and processes — to generate code needed for product and service development. Fully integrated into the company’s internal software development environment, the coding assistant supports natural language interactions and various programming languages to streamline developer workflows.

The Samsung Gauss Portal, a conversational AI service, is widely used for workplace tasks such as document summarization and translation, email drafting and more. Future updates aim to enhance the portal’s natural language question-and-answer capabilities and expand its multimodal features to include interpreting tables and charts and generating images.

Strengthening AI Capabilities Through Consumer Engagement

Leveraging its extensive product lines, Samsung is developing technologies that support on-device, cloud and generative AI to deliver a distinct competitive edge in AI capabilities.

By connecting with consumers through its products, the company gains valuable insights into user behaviors and needs to enable the development of practical AI solutions. This creates a virtuous cycle — whereby user data informs AI advancements, which in turn deliver value back to consumers.

Samsung couples this understanding of products with in-house advancements and strategic partnerships to implement optimal AI technologies. For instance, neural processing units (NPUs) — specialized accelerators for AI computation — must be tailored to meet the specific requirements of individual products, rather than applied uniformly across all devices. Samsung not only designs NPUs in-house for its own products but also applies software optimizations to partner chipsets, ensuring support for AI functions and peak performance in Samsung devices.

In addition, the company draws on the expertise and perspectives of product-specific specialists to address security issues and facilitate the immediate application of solutions to products while driving continuous technological improvement. The blockchain-based, multi-device security solution Knox Matrix is a notable example of how this collaborative approach can create an environment in which connected devices check each other for threats to enhance security.

▲ Samsung is leading in AI thanks to consumer insights gained from various products.

Through the implementation of customized AI technologies, Samsung continues to unlock new possibilities for users. Looking ahead, the company plans to deliver more differentiated AI experiences that empower users through meaningful and enriching interactions.

¹ Availability of Galaxy AI features may vary by region, country, OS or One UI version, model and carrier.
² Generative Wallpaper feature may vary by region and model.
³ Capable of simultaneously understanding and processing various types of data (modalities) including text, images, audio and video.

Samsung Electronics today hosted the Samsung Developer Conference Korea 2024 (SDC24 Korea), a virtual event that showcased the company’s latest software innovations and vision for the future.

Since 2014, Samsung Electronics has held this annual event to engage and collaborate with software developers, making SDC24 Korea the 11th iteration. This year’s conference highlighted research related to software embedded in products such as generative AI, software platforms, IoT, healthcare, communications and data. It also delved into the culture of open-source development.

During the keynote address, the company unveiled Samsung Gauss2, the second generation of its proprietary AI model, and highlighted its improved performance, efficiency and various application possibilities.

“Samsung Electronics is committed to developing cutting-edge software, including AI and data analytics, to enhance user experiences,” said Paul Kyungwhoon Cheun, President, CTO of the Device eXperience (DX) Division and Head of Samsung Research. “With three distinct models, Samsung Gauss2 is already boosting our internal productivity, and we plan to integrate it into products to deliver higher levels of convenience and personalization.”

Samsung Gauss2: A Multimodal Language Model Catering to Diverse Needs

Samsung Gauss2 is the follow-up to the company’s internal generative AI model unveiled last year, and it offers improved performance and efficiency in simultaneously handling various data types as a multimodal model integrating language, code and images. It is available in three distinct models catered to different purposes: Compact, Balanced and Supreme.

The Compact model is a small-sized model specifically designed to work efficiently, even in limited computing environments. It delivers optimized performance for on-device usage by maximizing the utilization of the device’s computing resources. On the other hand, the Balanced model focuses on achieving balance between performance, speed and efficiency, providing consistent performance across diverse tasks. Finally, the Supreme model aims for top-notch performance by leveraging Mixture of Experts technology¹ atop the Balanced model, allowing significant reductions in computational costs during training and inference processes while maintaining high levels of both performance and efficiency.

Samsung Gauss2 supports 9 to 14 languages as well as various programming languages, depending on the model. Samsung has developed and incorporated its own stabilization techniques for training large language models (LLMs) and designed a custom tokenizer² to ensure maximum efficiency for these supported languages.

The Balanced and Supreme models offer equal or superior performance in key metrics such as generating responses in various languages including English and Korean, and coding capabilities compared to leading open-source generative AI models currently available. Additionally, their processing speed per hour is 1.5 to 3 times faster, enabling quicker AI response generation, reduced user wait times and more efficient task handling.

Customizable for Various Productivity Tasks and Product Development

Having an internal generative AI model allows easier customization so that optimal performance can be achieved for specific goals and applications. Samsung Gauss is already being used in various tasks for Samsung’s employees, leveraging its customizable development capabilities.

With the power of Samsung Gauss, the in-house coding assistant ‘code.i’ assists the company’s software developers. Now upgraded to Samsung Gauss2, it is being utilized by business units within the Device eXperience (DX) Division and overseas research institutes.

Since its launch last December, the monthly usage of code.i has quadrupled, with about 60% of all software developers in the DX Division are now using it. Additionally, Samsung Gauss Portal is a conversational AI service powered by Samsung Gauss that assists employees within the DX Division as they handle various office tasks — such as document summarization, translation and email composition. The service was expanded to overseas subsidiaries in April. In addition, starting from August, Samsung is using Samsung Gauss to call center staff to automatically categorize and summarize customer calls.

Moving forward, Samsung aims to continuously increase productivity within the company by applying Samsung Gauss2, to improve the performance of the code.i service, enhance the natural language question-and-answer function of the Samsung Gauss Portal, and support multi-modal functions like understanding tables and charts and creating images.

Moreover, under the AI vision of “AI for All,” Samsung will continue to expand the reach of its AI-based services across all product lines so that users can experience more convenient and enjoyable daily lives. And by integrating knowledge graph technology with AI, Samsung expects to provide even more enhanced personalization services.

Presentations on Topics Including Software Platforms, IoT, Healthcare, Communications and Data

Following the announcement of Samsung Gauss2, there were presentations on the customer experience with Samsung’s platforms, which included relevant insights on the SmartThings experience and improvements to various software platforms.

Subsequently, 29 diverse technical sessions were conducted, including the following:

• The Future of Healthcare and Samsung’s Health Ecosystem Strategy
• SmartThings Customer VOC Experience Improvement with Generative AI
• code.i: Understanding Samsung Electronics’ AI Coding Assistant
• TV-based Lifestyle Content Hub To Enrich Your Everyday Experience
• AI Solution for Samsung Home Appliances Through AI Vision Technology and Data Utilization

For more detailed information about SDC24 Korea, visit the official website (https://www.sdc-korea.com/).

¹ An approach where only the most suitable expert models are selected and activated for specific tasks, allowing efficient performance enhancement through the conservation of necessary computational resources.
² A tool that converts inputs such as letters into data formats that can be processed by computers.