LG AI Graduate School to launch doctoral program

LG AI Graduate School has received approval to launch a doctoral program, expanding its offerings beyond the master’s program authorized in August.

The school, supported by South Korea’s LG Group and accredited by the Ministry of Education, will enroll 25 master’s students and five doctoral students each year.

The master’s course lasts three semesters, while the doctoral program will take at least three years and requires candidates to publish at least one SCI-level paper or present research at an international academic conference.

LG AI Graduate School will also operate joint programs with universities such as Seoul National University and KAIST, focusing on generative AI and large-scale foundation models for physical AI applications.

The doctoral program is set to begin with an opening ceremony scheduled for March.

🔗 Source: The Korea Herald

🧠 Food for thought

Implications, context, and why it matters.

LG’s doctoral track can aid EXAONE if GPUs and faculty measure up

Focus on generative AI and physical AI fits the EXAONE plan. EXAONE is LG’s multimodal foundation model with language and vision 1. The SCI-level publication bar or top conference talks matches frontier AI norms.
GPU access stays unclear. EXAONE 3.0 used 4×10^23 floating point operations per second (FLOPS) on NVIDIA H100s on Google Cloud Platform 2, and models like EXAONE Deep 32B with roughly 32 billion parameters take heavy resources 3. Without stated allocations or ties to SK Telecom’s NVIDIA H100 or B200 GPU as a Service setup the program may lag 4.
Links with Seoul National University and KAIST could widen research networks plus computing options. The Article gives no detail on operations or resource sharing.

AI infrastructure vendors can target the next South Korean doctoral intake

GPU clouds can serve five yearly LG AI Graduate School slots. SK Telecom’s Haein cluster has over 1,000 NVIDIA B200 GPUs 4, while the SK and Amazon Web Services build targets 60,000 GPUs by 2029 for academic tiers 5.
MLOps and experiment tracking can tailor for Korean-language work, given EXAONE’s bilingual scope 2. The three-year track starts in March, so vendors can time offers to procurement cycles.
Robotics vendors should prep for physical AI needs. LG applies foundation models in manufacturing and research, so projects will need simulators plus robot learning stacks such as Isaac Sim or MuJoCo for Korean labs 6.