Samsung Taps Stanford Scholar to Lead North America AI Research

* Editor's note: AI PRISM (Personalized Report & Insight Summarizing Media) is an artificial intelligence-based personalized news recommendation and summary service developed with support from the Korea Press Foundation. It selects and provides six news items tailored to each reader type.

1. Samsung Taps Stanford Scholar to Head North America Research, Shifting to Physical AI

Key summary: Samsung Electronics last month appointed Juan Carlos Niebles, director of Stanford University's Vision and Learning Lab and an adjunct professor of computer science, as head of the North America AI Center at Samsung Research at the executive vice president level. Niebles previously served as associate director of the Stanford-Toyota AI research center and as director of AI research at Salesforce, and has published more than 100 papers in physical AI-related fields such as computer vision, multimodal learning and robot control. With the appointment, Samsung Electronics is focusing on advancing physical AI businesses including humanoids, manufacturing AI and wearables, while also moving in earnest into related businesses, including the creation in July of a dedicated robotics organization in Korea called the RX Business Promotion Office. As competition in physical AI spreads across Big Tech, demand for talent in the robotics-AI convergence field is expected to expand rapidly.

2. Caught Using ChatGPT During an Exam and Given an F — Programs to Detect Cheating Emerge

Key summary: From last year through the first half of this year, a total of five cases of exam cheating using generative AI were identified at six major universities including Seoul National University, Yonsei University and Sungkyunkwan University — three at Sungkyunkwan, one at Seoul National and one at Yonsei. All the students caught at Sungkyunkwan received an F in the course, and at Yonsei, 40 students who came forward in a course with about 600 enrollees were suspended for two to three weeks. An AI cheating detection program developed by a Yonsei professor has drawn a series of requests for use from other universities, and the Ministry of Education drew up ethical guidelines on AI use on the 24th of last month, explicitly designating the copying and submission of AI answers during exams as serious misconduct. As institutional sanctions against AI cheating are strengthened across universities, observers say it is important to understand precisely what each school permits in terms of AI use.

3. SK hynix Launches SK hynix Ventures in Silicon Valley

Key summary: SK hynix held its first SK hynix Ventures Day in Silicon Valley on the 17th and officially launched its corporate venture capital brand, SK hynix Ventures. SK hynix, which has operated a CVC through its Americas unit since 2015 and has generated returns of more than twice its cumulative investment to date, used the launch to broaden its investment scope from semiconductors and future technologies to AI infrastructure as a whole, including AI computing, data centers, system software and optical interconnects. As demand for computing performance and memory capacity grows rapidly with the advancement of AI models and the importance of data center power efficiency increases, the company plans to further strengthen technology cooperation with related startups, SK hynix CEO Kwak Noh-jung said. Analysts say the expansion of the AI infrastructure ecosystem will widen employment opportunities in the semiconductor, software and data center fields.

4. Xenon Open-Sources Computer-Operating AI Model Hunmin VLM 397B

Key summary: Generative AI solutions company Xenon has released as open source a computer-operating AI model, Hunmin VLM 397B, with 397 billion parameters. The model identifies locations to manipulate on a screen, such as buttons and input fields, to carry out actual computer tasks. It scored 75.6 points for accuracy on ScreenSpot Pro, a high-difficulty screen recognition benchmark, ranking second on the official Hugging Face leaderboard and standing as the only model developed by a Korean company among the 48 models currently listed. On OSWorld, a benchmark for tasks on the Linux operating system, it scored 70.5 points, 22.3 points higher than the base model, and it completed the entire post-training process using only eight Nvidia B200 GPUs, demonstrating technical efficiency in achieving high performance with limited computing resources. Amid expectations that the scope of agentic AI applied to actual corporate work will gradually expand, Korean AI software technology is drawing attention at the global level.