Reflecting recent trends, the ‘2024 Artificial Intelligence & Big Data Show’, Korea’s largest artificial intelligence (AI) exhibition held at COEX in Seoul until the twenty first, was dominated by large language model (LLM) and data specialized corporations.
Wisenut (CEO Kang Yong-seong), with over 20 years of experience, is a representative example. As now we have been carrying out numerous collaborative projects with public institutions, we also introduced numerous related solutions at this exhibition. A very good example is ‘Dtokbi,’ a chatbot built to enhance the work efficiency of Hwaseong City’s civil service counselors.
The reason is that it minimizes hallucinations through technologies resembling Search Augmented Generation (RAG) and alleviates counselors’ difficulties by finding relevant documents in response to the content of the criticism.
Conan Technology, which supports multimodal AI search and document creation with its own LLM, can be a flagship company.
Recently, we’re specializing in constructing specialized models for industries resembling the defense sector.
Visitors also flocked to the Testworks (CEO Seok-won Yoon) booth.
Particularly, the camera-equipped handheld remote control robot ‘CEDI’, which serves as a multi-modal data collection platform, goals to gather high-quality data for indoor and outdoor autonomous driving.
He said that for future autonomous driving technology, data should be collected not only in urban areas but additionally in indoor locations resembling airports, ports, warehouses, shopping malls, and manufacturing plants.
Allganize (CEO Lee Chang-soo) has recently been showing strength in industry-specific models resembling the finance-specialized LLM leaderboard.
While announcing the launch and development of a small language model (sLM) in the longer term, this time they introduced a brand new model, ‘Alpha-Instruction’.
He predicted, “As we’re strong in Korean language skills, we’ll receive many contacts from domestic corporations.”
Reporter Jang Se-min semim99@aitimes.com
