Generative artificial intelligence service products such as ChatGPT and Midjourney are attracting the attention of the market. Against this backdrop, the Korea Artificial Intelligence Industry Association (KAIIA) held the ‘Gen-AI Summit 2023′ at COEX in Samseong-dong, Seoul. The two-day event aims to promote and advance the development of generative artificial intelligence (AI), which is expanding the entire market.
On the first day, starting with the keynote speech by Jin Junhe, head of the artificial intelligence fusion business department, large technology companies such as Microsoft, Google and AWS actively developing and serving ChatGPT, as well as fabless industries developing artificial intelligence semiconductors attended and made relevant Presentations, including “NLP Changes Brought by ChatGPT” by Persona AI CEO Yoo Seung-jae, and “Building a High-Performance, Power-Efficient and Scalable AI Inference Chip for ChatGPT” by Furiosa AI CEO Baek Jun-ho .
Jin Junhe said that in 2023, the year of the artificial intelligence war, the ChatGPT plug will enter the market as a new game rule for the huge language model competition between Google and MS. In this case, he foresees opportunities in AI semiconductors and accelerators that support AI models.
Furiosa AI is a representative fabless company manufacturing AI semiconductors in Korea. Furiosa AI CEO Baek, who is working hard to develop general-purpose AI semiconductors to catch up with Nvidia, which holds most of the world’s market in hyperscale AI, is convinced that “the demand for chips in the AI field will explode in the future”
As AI services become more complex, they inevitably face increased infrastructure costs. Nvidia’s current A100 and H100 GPU products have the high performance and computing power required for artificial intelligence computing, but due to the increase in total costs, such as high power consumption and deployment costs, even ultra-large-scale enterprises are wary of switching to next-generation products. The cost-benefit ratio expressed concern.
In this regard, Baek predicted the direction of technological development, saying that in addition to more and more companies adopting artificial intelligence solutions, the market demand will be to maximize efficiency and performance within a specific system, such as “energy saving”.
In addition, he emphasized that the spread point of artificial intelligence semiconductor development in China is ‘usability’, and said how to solve development environment support and ‘programmability’ will be the key.
Nvidia has built CUDA to showcase its support ecosystem, and ensuring that the development community supports representative frameworks for deep learning such as TensorFlow and Pytoch is becoming an important survival strategy for productization.
Post time: May-29-2023
