Today, South Korean electronics giant Samsung made a significant foray into the generative AI space with the announcement of Gauss, a foundation model designed to run locally on smartphones and produce text, code, and images. The company unveiled this groundbreaking development at its AI Forum in Suwon, where it showcased its efforts and highlighted the internal testing being carried out by its employees.
Gauss, named after the renowned German mathematician and physicist Carl Friedrich Gauss, who pioneered the theory of normal distribution, serves as the basis for modern machine learning and AI. The ultimate goal for Samsung is to enhance and evolve Gauss for various product applications, creating novel user experiences.
The Three Versions of Gauss: Language, Code, and Image
While Gauss is still in its nascent stage, Samsung Research has confirmed that it will consist of three versions: Gauss Language, Gauss Code, and Gauss Image. The language model, similar to Google Workspace’s generative AI capabilities, will assist users in tasks like composing emails, summarizing documents, and translating content. Additionally, Samsung hints at the possibility of leveraging Gauss Language for smarter device control, although specific details have not been disclosed.
Gauss Image, on the other hand, will cater to photo editing and enhancement on Samsung devices. Its features will range from generating and editing images to adding elements and improving resolution. Essentially, it will grant users access to generative-fill-like functionalities right within their smartphone’s photo editor.
The final offering, Gauss Code, is designed to serve as a software development assistant, facilitating efficient code writing for teams. By providing functions like code descriptions and interactive interface-enabled test case generation, Gauss Code streamlines the development process.
The Implications of Integrating Generative AI for Samsung Users
Integrating generative AI into the Samsung ecosystem holds the promise of a significant upgrade for the company’s customers. While no official timeline has been announced, Samsung maintains that it is currently using the Gauss model to enhance employee productivity. However, it plans to expand its usage to diverse product applications in the near future, hinting at a potential integration in its flagship devices as early as 2024.
This timeline aligns with Qualcomm’s plans to launch its next-generation chip equipped with an AI engine, supporting multi-modal generative AI models, large language models, language vision models, and transformer network-based automatic speech recognition. Since Qualcomm is Samsung’s vendor for mobile SoCs, this development intensifies the competition in the race for on-device AI, a domain also being explored by companies like Google and Apple.
Unlike cloud-based general-purpose models, dedicated hardware and AI models running on devices can deliver superior results, especially when accessing device-specific data like driving patterns, restaurant searches, and photos. As a result, users can expect more personalized outcomes, surpassing the capabilities of existing AI solutions.
To ensure the successful implementation of its AI vision, Samsung has established an AI Red Team tasked with identifying and resolving any security and privacy concerns that may arise. In the coming months, the company is expected to share further details about its initiatives in this area.
