SK Telecom is developing technology to improve inference performance and computational efficiency for its own AI foundation model through joint research with the Massachusetts Institute of Technology (MIT) in the United States, according to the SKT Newsroom on Wednesday. MIT professor Kim Yoon-hyung, who is conducting joint research with SKT on large language model performance improvement technology using test-time training (TTT), said talent and experience operating large-scale systems are core to sovereign AI competitiveness. “Computing resources can be bought with money and data centers can be built, but experience and capabilities from directly operating large-scale systems take much longer to accumulate,” Kim said. He noted that AI performance depends on scaling relatively simple algorithms, making it important to secure researchers and engineers who can directly handle large systems. Kim said that as AI has become a key factor in national security, it is natural for countries to build their own AI ecosystems, but confining sovereign AI goals too much within a “sovereignty” framework could spur excessive competition rather than cooperation. TTT is a technique where a deployed AI model performs brief additional learning based on relevant material before answering a specific question, using part of its computing resources during inference to refine performance. Kim explained that performance can improve even if additional computing resources are invested in the inference stage. The research also focuses on helping AI agents accurately understand long contexts at low cost.