[AI Leadership] ② Shaping Personalised AI With Data Intelligence and Device Connectivity

The age of personalised AI is here. Samsung Electronics offers safe, tailored AI experiences backed by user insights across its products — from mobile devices and TVs to home appliances — through continuous innovation in software and platform technology.

Just as performance metrics and physical measurements are essential to designing customised workout routines for professional athletes, data plays a crucial role in creating personalised AI. As such, Samsung is prioritising data intelligence research to transform vast amounts of data into meaningful and intuitive user experiences.

Samsung Newsroom takes a closer look at how Samsung’s advanced data research and innovative connectivity solutions are pioneering the era of personalised AI.

The Power of Data Intelligence in Personalised AI

A truly personalised AI experience starts with understanding the user. Data intelligence identifies a user’s needs at specific moments based not only on historical device usage data but also on the user’s present context. This enables appropriate services to be delivered in a timely manner. However, there are challenges — for example, users don’t hand in refined reports of their needs and wants; the matching of expected needs and potential services must be carried out in real time; and most of all, personal information must constantly be kept safe.

As Samsung reaches consumers across nearly all daily touchpoints — including work, entertainment, household chores, health and security — the company has the unique opportunity to gain deep insights into user behaviors and preferences.

In order to harness this potential, Samsung has strengthened its data technology capabilities by researching user data collected annually from hundreds of millions of devices and services — all in strict compliance with internal and external regulations. As AI models develop a better understanding of users through data collection, processing and analysis, they can leverage retrieval-augmented generation (RAG) to deliver more personalised services for different contexts. By using technologies to connect reliable, logic-based knowledge bases to large language models (LLMs), these AI models access the latest, most reliable information and find the most suitable answers.

As Samsung focuses on technologies that connect data to refine personalised AI experiences, a key question is how can such vast and complex data be managed more effectively and tailored precisely to individual needs? The answer, in part, lies in knowledge graphs.
 

Enhancing AI Experiences With Knowledge Graphs

A knowledge graph is a knowledge representation technology that organises related information in an interconnected graph format, making it easily interpretable for both humans and computers. By defining individual entities within the data and linking them, their relationship can be represented as a “triple” in the form of “subject — predicate — object.” For example, “the capital of South Korea is Seoul” can be represented in the following graph, “South Korea — capital — Seoul.” Various types of knowledge such as people, objects, times and places can be structured in a highly refined manner by connecting numerous entities in this triple format.

Unlike traditional relational databases that fill tables with information, knowledge graphs effectively store data and quickly retrieve information or new inferences as needed.

When combined with Samsung’s on-device AI, knowledge graphs allow users to experience highly customised AI services while protecting sensitive personal data from external exposure. This technology is expected to be applied across various products and devices.

The key to implementing knowledge graphs is the ability to quickly find specific answers from information represented in the graph. Converting constantly changing real-world data into graphs and efficiently extracting information requires complex computations — and as the graphs grow, rapid information retrieval becomes more challenging.

To bolster its capabilities in this area, Samsung acquired Oxford Semantic Technologies in July. Oxford Semantic Technologies, a U.K.-based startup renowned for its knowledge graph technology, specialises in optimising data processing and advanced inference to facilitate swift understanding and precise extraction of information from knowledge graphs.

Samsung has collaborated with Oxford Semantic Technologies on graph databases and ontologies^[1] since 2018, and the recent acquisition is anticipated to create synergy in personalised AI research across Samsung’s product ecosystem. The company plans to develop applications that offer valuable experiences to consumers and explore new research avenues in various areas including the integration of generative AI and knowledge graphs.

Empowering Everyday Life With Data-Driven Insights

Samsung is rapidly incorporating personalised AI features powered by data intelligence across a wide range of products and services to enrich users’ daily lives.

Digital healthcare is one example. The Galaxy Watch7 and Galaxy Ring comprehensively analyse user health data including sleep, exercise and heart rate to offer an Energy Score via the Samsung Health app. Going beyond simple data measurement, this analysis provides an easy-to-understand numeric summary of a user’s daily health status and personalised insights for individual activities. By researching the weighted importance of each factor in the Energy Score based on age and gender, Samsung has developed a more customised approach to health management.

The entertainment experience has also been enhanced through data-driven insights. In the U.S., Gaming Hub — Samsung’s game platform available on its smartphones — uses explainable AI (XAI) to offer personalised game recommendations. Leveraging a model of the user’s preferences, this technology not only recommends games but also explains the reasoning behind each suggestion to help users select the games that suit them best.

Refined AI Experiences With SmartThings

Personalised AI will evolve as more devices connect with each other and acquire usage data from users’ daily lives. At Samsung Developer Conference 2024 (SDC24) in October, Samsung unveiled Home Insight — a service that allows users to manage and control their smart home through SmartThings. The feature analyses users’ lifestyle patterns, device usage history and home conditions to provide real-time reports and tailored recommendations such as turning off devices or utilising specific features based on different situations.

In addition, Samsung has a personalised AI feature to support the well-being and safety of family members in need. Launched in June, SmartThings Family Care allows users to remotely monitor family members and take any necessary actions through SmartThings-connected devices and location-based data. The company plans to expand these AI-connected capabilities to health devices, supporting family health in areas like sleep and diet.

Samsung is advancing technology that allows devices to recognise users and provide seamless, personalised experiences across all connected devices. For example, the company is planning to enable users who have large text mode activated on their smartphones to automatically apply the setting to other devices with a single voice command. Additionally, Samsung is exploring functions that use sensors and devices within the home to detect a user’s location and adjust lighting, temperature and humidity according to individual preferences.

Personalised AI Needs Robust Security

Security is a fundamental requirement for delivering safe, personalised AI. To ensure the protection of personal information and address biases in AI systems, a multi-faceted approach is needed — including improving data processing and model training processes and implementing a rigorous verification process for outputs.

When collecting internal and external data for AI model training, Samsung performs data pre-processing to check and filter for privacy violations. Additionally, the company is strengthening its safety verification of AI outputs by addressing risks related to harm, bias, privacy and jailbreak^[2] vulnerabilities.

Samsung’s Knox Matrix further elevates security in a multi-device connected environment. Using blockchain technology, connected devices can monitor each other for security threats through Trust Chain; securely share user data by transmitting only encrypted information to servers via Credential Sync; and apply consistent security standards across operating systems such as Android, Tizen and Windows with Cross Platform. Samsung plans to expand Knox Matrix to cover its mobile devices, TVs and home appliances.

 
^[1]A search method that finds objects by adding logic such as connection relationships and inclusion relationships.
^[2]Attack methods that bypass safety measures and ethical guidelines of AI models to create malicious content.