On-device AI, e-commerce innovation, and AI for water management
Hankyung
Nov 28, 2024
On-device AI, LLM-based e-commerce solutions, and smart water management are some of the services offered by promising domestic AI startups that participated in the 65th AI Future Forum Startup Roundtable held recently. The event was hosted by Hankyung Geeks, the 'Citizens' Coalition for Scientific Society' (CCSS), AI Future Forum (AIFF), and Capstone Partners to connect companies and investors.
ZETIC.ai: Developing integrated on-device AI software
ZETIC.ai is an on-device AI software startup. Yeonseok Kim, CEO of ZETIC.ai, stated, "According to OpenAI's announcement in September, OpenAI's revenue is 5 trillion KRW, but expenses amount to 12 trillion KRW, resulting in a loss of 7 trillion KRW." He added, "Most of these expenses come from GPU cloud costs." Kim further explained, "GPUs themselves are expensive, but the infrastructure and maintenance costs are also extremely high."
Kim continued, "Given this situation, several efforts have been made to reduce related costs, and on-device AI, where AI runs directly on the user's device, has begun to gain attention." He added, "Since the AI model must run on the device, the model needs to be small, and software to run such models is also required."
Kim also mentioned, "Recently, the NPU (Neural Processing Unit) in smartphones is up to 60 times faster compared to traditional CPUs and GPUs, while also consuming more than 90% less power." He pointed out, "Unlike the GPU market, there are no major players like NVIDIA in the mobile NPU market." He added, "One of the barriers to entry is that each manufacturer uses different NPUs, making it impossible to use a single software for all devices."
Kim noted, "In this environment, some companies have given up using NPUs and instead focused on using CPUs and GPUs while developing smaller AI models. However, reducing model size may speed up computation but can sometimes lead to reduced model performance." ZETIC.ai is developing software to address these challenges.
In July, ZETIC.ai launched a trial version of its on-device AI integrated solution, 'ZETIC.MLange.' It is an automated solution that transforms existing AI models into on-target AI model libraries for integration into on-device AI systems. Melange allows users to create on-device AI applications easily, quickly, and conveniently by utilizing their AI models. It supports various NPUs from different SoC manufacturers and different operating systems, enabling on-device AI regardless of the device. Kim revealed, "We are also expanding our overseas business, discussing the adoption of our solution with a healthcare company in India."
Corka: Supporting e-commerce platforms with AI
Corka is an AI SaaS all-in-one marketing solution company. Corka's CEO, Younghyun Jeong, said, "We primarily focus on large language models (LLMs) and recommendation systems," adding, "In the retail media platform domain, we analyze real-time user behavior data to advertise or recommend products that consumers are more likely to click on and purchase, thereby helping e-commerce platforms increase their revenue."
Jeong further elaborated, "We are also developing other LLM-based solutions that provide value to users. One such solution is 'Moonlight,' an AI-based PDF document analysis tool." Moonlight is an AI academic paper analysis solution that easily summarizes and translates PDF versions of papers, incorporating a conversational feature based on analysis data. It also offers features like smart citation (showing relationships between cited documents and the user's library) and automatic classification/indexing of libraries (industry-specific indexing).
Jeong added, "We are continually launching AI services that revolutionize work environments, such as the conversational AI calendar 'Trace.' We are integrating our capabilities from LLM into e-commerce platforms and are currently in talks with related companies, expecting revenue from this next year." He emphasized, "Advertising and recommendation are the most profitable areas for AI, and platforms like Google and Facebook generate a significant portion of their revenue from ads." Jeong also mentioned that they are collaborating with domestic law firms to utilize LLM capabilities in security-related applications.
Ecopiece: AI robots for water management
Ecopiece is a smart water management solution startup. Ecopiece CEO Inwon Chae emphasized, "The future requires not only advanced high-tech solutions but also effective water management in response to climate change." He explained, "The water humans use is highly sensitive in terms of quality, and it is currently stored and used, with issues like algal blooms occurring at the Soyang River Dam since last year." He added, "The semiconductor industry also requires large amounts of water, and there are predictions that production costs will significantly rise due to water shortages."
Chae noted, "The water resource industry ecosystem is highly analog, with humans making decisions and adding chemicals manually, which is vulnerable compared to data-driven AI." He explained, "In cases of algal blooms, the current approach is for people to navigate in boats to remove or apply chemicals, which is inefficient, costly to maintain, and potentially dangerous." Chae highlighted, "Ecopiece's differentiation lies in its low maintenance cost, unmanned automation, and 3D water quality monitoring."
Ecopiece has developed 'Ecobot,' an AI autonomous waterborne robot that can freely navigate the water, measure water quality, analyze sources of pollution, purify water, and prevent the spread of water pollution. Ecobot is a solution for managing water resources in facilities such as reservoirs, urban lake parks, and agricultural ponds through waterborne robots rather than humans. It can detect and remove algae when it occurs and monitor pollution levels in real-time. The AI solution can also address various water quality issues, such as fish deaths due to rising water temperatures, odor generation, and turbidity increases.
Chae added, "Initially, we collected data with autonomous robots, and as the business expanded, we launched solutions for pollution purification." He emphasized, "The robots can operate on water 365 days a year without recharging, solving energy-related issues." Looking ahead, he revealed, "We plan to strengthen our B2C business by offering a futuristic waterborne leisure solution where people can enjoy riding on the water quality management robot."