Generative AI may help resolve chip industry challenges

Samsung Electronics President Kyung Kye-hyun delivers an opening speech for the Samsung AI Forum. [Courtesy of Samsung Electronics]

Rapid advancements in generative artificial intelligence (AI) technology and its potential to address critical challenges were shared during this year’s Samsung AI Forum that took place on Tuesday in Suwon, Gyeonggi Province.

The forum, held under the theme “Large-scale AI for a Better Tomorrow,” featured a broad range of discussions on generative AI technologies with industry experts and innovators.

Samsung Electronics Co. President Kyung Kye-hyun referred to generative AI technology as “an innovative means to address the challenges faced by humanity,” adding that it will also “help resolve some of the most complex problems in the semiconductor industry.”

Kyung further emphasized the company’s commitment to playing a pivotal role in enhancing the AI ecosystem through AI computing system components, including high-bandwidth memory chips.

He also highlighted “safety, reliability, and sustainability” as remaining challenges in large language models and other generative AI technologies, adding that these AI technologies will need to address hallucination issues and improve operational efficiency.

Industry innovators and academics also shared insights at the forum, including Jim Keller, chief executive officer of AI chip company Tenstorrent Inc.

Keller, in his speech, emphasized the importance of open-source approaches while proposing ideas to overcome the limitations of AI technology through chip design innovation.

He advocated for innovation in hardware structure based on open instruction set architecture, such as Reduced Instruction Set Computer Five (RISC-V).

Keller believes that investing in AI chip design complexity can be reduced through developing an open ecosystem.

Keller also expressed trust in Samsung Electronics, referring to the South Korean electronics maker as the best partner for AI chiplets.

Tenstorrent recently named Samsung Electronics as its chiplet partner, with plans to produce AI chiplets using Samsung Electronics’ 4-nanometer process, SF4X.

Chiplets involve packaging multiple semiconductors into one package.

Yoshua Bengio, professor of computer science at the University of Montreal, also introduced AI algorithms that can prevent AI technology, based on large language models (LLM), from deviating from its intended purpose.

Researchers from the Samsung Advanced Institute of Technology (SAIT) also actively engaged in the discussion, providing various insights.

During the forum, the Samsung AI Researcher Awards ceremony was held, awarding five individuals, including Jason Lee, professor of electrical engineering and computer science at Princeton University.

Lee was recognized for his contributions to AI theory and applied research, focusing on deep learning, reinforcement learning, and optimization.

The seventh annual Samsung AI Forum has seen over 1,000 participants this year. The second day of the forum will be held on Wednesday by Samsung Research, the research arm of Samsung Electronics, at its Seoul campus.