“In 10 years, we will be wondering whether we should award AI not only the Nobel Prize in Science but also the Nobel Prize in Literature” (Nam Se-dong, CEO of VoyagerX)

"AI is still in its infancy. The value that deep learning will create over the next 10 years will be enormous." (Kim Jong-yoon, CEO of Scatter Lab)

"After AI service development, talent across all fields, from development to HR and PR, will be needed. With experience, greater opportunities will arise for everyone." (Kim Tae-soo, CEO of Neosapience)

"The AI friend from the movie Her will be in its early stages" (Park Gyu-byeong, CEO of Tunip)

This is the current and 10-year outlook for AI as diagnosed by representatives of Korean AI startups at the ai.x 2021 conference hosted by SKT.

SK Telecom (CEO Park Jung-ho) held the ai.x 2021 conference under the theme of 'AI For Everyone'.

On the 15th, the current status and future outlook of Korea's representative startups were revealed, along with SKT's efforts to secure AI data, and on the 16th, a sub-theme called 'AI for X' was presented.

At the opening session on the 15th, hosted by B Factory CEO Noh Jeong-seok, candid discussions were held about the difficulties faced by the startup industry. />
For AI big models, which are difficult for startups to achieve, he predicted, "If we secure the technology and receive significant value investment, it will be possible." He also expressed hope for collaboration with large corporations, saying, "The (technical) threshold will gradually lower, and collaboration opportunities will increase." He also expressed confidence as a startup, saying, "There are many businesses that can be pursued with models of meaningful size."

SKT developers also held an in-depth discussion on securing and utilizing the vast amounts of data that form the foundation for AI development. The developers shared their experiences securing high-quality big data in diverse fields, including healthcare, language, and transportation.

“If there is a problem with the reliability of the data after going through the process of comparing and contrasting the collected data, we find errors in the process of collecting and transmitting the data and use the corrected data,” he said, revealing a method for securing high-quality data.

He also revealed the challenges of data processing and purification. "(In the case of language data) personal information requires multiple reviews, and new profanity and hate speech are constantly being created, making it difficult to replace with sanitized expressions. Purifying this will remain a significant challenge going forward," he said, emphasizing efforts to secure high-quality data.

Additionally, a plan was proposed to enable external use of the high-quality data that was obtained with difficulty. An SKT expert stated, “We plan to provide in-house data as valuable content by area through a service (tentative name) called Puzzle in the future, and share publicly available data externally in the form of APIs and services.”

A discussion was also held with experts from NHN, CAST.ERA, and SK Hynix, the partners of AI semiconductor 'Sapion', which is evaluated to have lowered the entry barrier for AI infrastructure construction through SKT's own technology development.

NHN's Director Kim Joo-hwan commented, "We are currently implementing a government-funded demonstration project by deploying Sapion in our own data center, and with SKT's active technical support, the project is proceeding smoothly." NHN stated that through future demonstration projects, they plan to verify and commercialize services such as video and image quality improvement services utilizing Supernova technology, an indoor image positioning system, and bone age analysis using X-ray data.

Kevin Gage, COO of Cast.era, the broadcasting group of Sinclair Broadcasting, the largest terrestrial broadcaster in North America, expressed his expectations, saying, “Sapion is very efficient and consumes less power than existing GPUs in AI inference, so it is possible to provide video services that convert to high-definition in real time even with small infrastructure.” He added, “With the next version of Sapion, Sinclair’s large-scale real-time AI broadcasting in the U.S. will be possible at low cost and with low power.”

On the second day, the 16th, sessions will be held on industrial AI, bio AI, safety AI, and AI application cases. It is expected that the day will be hosted by non-experts, allowing the general public to ask a variety of questions about AI.

During the conference period (September 15th-16th), you can participate in real time on SKT's AI homepage (aix.skt.ai), and from the 17th, you can 'watch again' without logging in.