I have maintained the position that open source will win since the past. Even OpenAI, which has insisted on a closed model, recently announced that it will soon launch an open source model. As the pressure for a shift to open source intensifies within the industry, there is no choice but to release an AI platform that is accessible to everyone.
Ilya Polosukin, founder and CEO of Near Protocol, noted this on the 11th at a press conference held at Sky31 Convention in Songpa District, Seoul, during the "Open Source AI Summit Seoul." The "Open Source AI Summit Seoul" is part of the conference "Beedle Asia Week," organized by the blockchain community builder corporation CryptoPlanet. The event included discussions aimed at creating an open source ecosystem, with participation from CEO Polosukin, Erica Kang, founder of CryptoPlanet, and Andu Kyung, CEO of Good Gang Labs.
Polosukin, originally from Ukraine, joined Google Research in 2014 and engaged in AI technology development and research. He is known for being a co-author of the paper "Transformer: Attention Is All You Need," published in 2017. This paper presented the "Transformer" for the first time, which has been adopted by many AI deep learning models, including ChatGPT. He later left Google and co-founded Near Protocol, a blockchain mainnet development company, in 2017.
CEO Polosukin emphasized the transition to an open source ecosystem for AI innovation. He said, "The difference between open source models and closed models can be thought of as user-owned AI versus corporation-owned AI." He added, "Corporation-owned AI is developed mainly with the goal of maximizing the company’s revenue, while user-owned AI is developed in the most useful way for individual developers."
The competition between open source and closed models, which has existed for a long time in the IT field, is now spreading to the AI sector. Open source refers to making source code, which can be considered as software blueprints, available for anyone to use. In contrast, closed models require users to pay an expense to use software owned by the development company. The competition that previously occurred between the two factions in developer computer operating systems and the mobile market is now shifting to the generative AI space. Initially, corporations like OpenAI and Google insisted on closed AI, but recently, companies like Meta and DeepSeek have intensified their pursuit of open source.
CEO Polosukin explained that blockchain could play an incentive role in the transition to open source. He stated, "As AI models are released monthly, the speed of innovation is crucial, and if we utilize blockchain as the main incentive tool for developers, innovation would be possible while maintaining confidentiality, securing data sovereignty and ownership."
He diagnosed the South Korean market as having "great opportunities," but noted that "language models and interaction models have many inherent biases, and there is a lack of data that reflects Korea's unique culture and current situation in real time." He further stated, "In Korea, especially in the entertainment sector, there is a leading position, and I believe South Korean startups can contribute to the global expansion of their native content by utilizing open source AI, which they have fine-tuned themselves."
The startup sector forecasted that domestic corporations could gain global competitiveness through open source AI. Andu Kyung, CEO of Good Gang Labs, remarked, "Open source AI is a necessity, not an option, for startups," and added, "Considering the infrastructure and funding that domestic startups have, it is difficult to create a large language model (LLM) without open source." He continued, "In South Korea, AI development specialized in Korean ultimately comes down to a data competition, and if large corporations can create an environment that allows them to freely disclose data while ensuring safety, I believe that domestic startups can have sufficient global competitiveness."