Artificial intelligence (AI) is entering a stage where it goes beyond information retrieval and Q&A to assist in semiconductor design and verification. Next-generation semiconductors such as high-bandwidth memory (HBM) are so complex in design that signal, power, and thermal issues must all be considered simultaneously. AI “agents” are now being introduced into such semiconductor design processes.

On the 5th, TERA Lab, led by Professor Kim Jung-ho of the School of Electrical Engineering at KAIST, announced that it will hold an online event, “TERA AI Agent Day,” on the 11th to unveil the AI agent ecosystem currently being applied to semiconductor design.

In his keynote speech, Professor Kim will present a blueprint for a “single engineer–multiple agents” model, in which one engineer directs numerous specialized AIs. Rather than entrusting all the specialized functions required for semiconductor design to a single AI model, the approach links task-specific agents with design software to solve complex problems. Human judgment is introduced at necessary stages.

Student researchers will present case studies applied in actual research. These include an agent that analyzes the signal integrity of HBM and next-generation memory high-bandwidth flash (HBF), and a system in which multiple AIs collaboratively optimize the placement of electrodes that connect stacked chips. TERA Lab aims to have each student develop at least two agents.

Professor Kim stated, “In the future, the competitiveness of engineers and developers will be determined by their ability to directly design specialized AI agents and enable them to cooperate organically,” adding, “Future researchers will evolve into ‘AI agent architects’ who design and operate AI agent ecosystems.”

Popular News