DeepSeek, which has recently

DeepSeek, which has recently emerged as a big topic in the AI industry, has made an innovative change to the existing LLM (Large Language Model) paradigm. Although the domestic stock market was closed for the Lunar New Year holiday, the global stock market is sensitive to the DeepSeek announcement, showing volatility centered on technology stocks. The shock from China, which can shake the market capitalization of the water level by hundreds of trillion won due to a shift in technical thinking at the beginning of the second Trump administration, is certain to be a story that will be consumed politically in the future.

In addition to existing open-source LLM models such as Meta's LLMA, LLM technology will become more common faster if powerful models such as DeepSeek are released as open-source.

DeepSeek introduces technological innovations such as "cross-modal learning" to show its ability to understand and process different data, including text, images, speech, and video. However, the low computational cost that many people note may not be as significant as you might think. The performance of AI models depends on how many variables are branched and how many possibilities are controlled and judged, and reducing variables and branches can naturally lead to exponential drops in computational costs.

For example, it's important what services Naver and Google compete with, but how quickly their websites use fewer resources is likely to be uplevel in the long run. Of course, this could pose a threat to companies like NVIDIA, which relies on parallel processor advancements, given that DeepSeek has demonstrated that it can create models with less computational power. The plunge in NVIDIA's stock price will not be unrelated.

In recent conversations with AI-related chip designers, there are many opinions that the computational cost of AI models will decrease exponentially in the future. This is similar to the cryptocurrency market's transition from GPU mining to ASIC mining, which has improved energy efficiency by more than 100 times. In AI, when the threshold is reached, the time when general-purpose hardware is converted to specialized hardware will surely come and come soon.

With LLM technology rapidly becoming commonplace, Korea has many concerns about what to look for. The release of models such as DeepSeek as open source means that what Korean developers will build up based on this is likely to be a key competitive edge in creating added value.

However, given the Yoon Suk Yeol government's populist and unfounded policy of expanding the number of medical schools along with the low birth rate, we are pessimistic about whether South Korea will be able to secure enough science and technology personnel to create this added value. Therefore, restoring the issue of expanding the number of medical schools and going all out to secure science and technology personnel should be a major political task.

In addition, Korea has a huge amount of digitized data through e-government. Millions of public officials produce data every day, and learning and utilizing them will be an important asset for enhancing national competitiveness. Social efficiency can be greatly increased by anonymizing judgments, investigation data, tax payment data, and educational data without worrying about personal information leakage. We need to be bold rather than being a frog in a well because we are afraid of maggots.

However, there are concerns about conflicts with professionals such as lawyers, doctors, pharmacists, and teachers in this process. It is the role of politics to coordinate and make a soft landing in these conflicts, and we need to think about how Korea will face this wave in the age of AI. Whether you will ride the AI tide or face the wave head on depends on future political discussions and decisions.

The New Reform Party wants to drastically change the paradigm and thesis of politics dealing with the future in a regulatory-oriented Korea. We will realize a politics that truly transforms our perception, not a politics that pretends to know by imitating a trendy Buzzword.

Watching foreign friends who used to study abroad writing and discussing on Facebook with enthusiasm on these topics, the current political situation in Korea, which has been hampered by conspiracy theories for fraudulent elections, is so cruel. I hope that the venue will be opened to reject conspiracy theorists as soon as possible and discuss important issues for the future of Korea.

In the frustration, I ran away from reading less readable articles for the first time in a long time.

Happy New Year, everyone.