[2024 Future Business Forum] Kim Donghwan, CEO of FortyTwoMaru, "High-Cost Mega AI... Need to Consider Cost as Well"

Keynote Speech at Asia Future Business Forum on 22nd
'AX Era Mega AI Utilization Strategy' Presentation

Kim Dong-hwan, CEO of FortyTwoMaru (photo), said, "When companies adopt ultra-large artificial intelligence (AI), they also need to consider the cost incurred per question."

Kim Dong-hwan, CEO of FortyTwoMaru, is delivering the keynote speech at the 2024 Asia Future Business Forum hosted by Asia Economy at Lotte Hotel in Jung-gu, Seoul on the 22nd. Photo by Kim Hyun-min kimhyun81@

On the 22nd, at the keynote speech of the ‘Asia Future Enterprise Forum’ held at Lotte Hotel in Sogong-dong, Seoul, CEO Kim pointed out the high cost issue as one of the considerations companies should keep in mind when introducing large language models (LLM). While presenting on the ‘AX (AI Transformation) Era Ultra-large AI Utilization Strategy,’ he highlighted hallucination, security, and cost issues as key points to consider. He added, "Although ChatGPT does not disclose its costs, I understand that currently, the fee charged to users is less than the cost per question. There may come a time when companies will have to pay the cost per question."

FortyTwoMaru is an AI startup developing a QA (Question Answering) platform that derives a single correct answer from vast unstructured data. It achieved joint first place with Google AI team in the MRC (Machine Reading Comprehension) competition hosted by Stanford University and also secured first place in the language understanding competition ‘GLGE’ organized by Microsoft (MS), gaining recognition for its technology in global competitions.

He mentioned Retrieval-Augmented Generation (RAG) technology as an element that can solve LLM’s problems and enable commercialization. RAG supplements models trained on existing data with customized databases to reflect the latest information, improving the model’s accuracy and quality. This helps reduce the hallucination phenomenon, a drawback of LLMs. FortyTwoMaru possesses its proprietary technology ‘RAG42,’ which extracts and refines only the optimal information required by LLM models and supports the development of domain-specialized engines for different industries.

Kim Dong-hwan, CEO of FortyTwoMaru, is delivering the keynote speech at the 2024 Asia Future Business Forum hosted by Asia Economy at Lotte Hotel in Jung-gu, Seoul, on the 22nd. Photo by Kim Hyun-min kimhyun81@

As another alternative to complement the drawbacks, he also mentioned small large language models (sLLM). FortyTwoMaru supports enterprise-specific modes with its proprietary technology ‘LLM42’ to mitigate hallucination in LLMs. FortyTwoMaru is commercializing its technology through businesses in QA, TA (Text Analytics), AI chatbots, and smart devices. It collaborates with LG Uplus, Naver Cloud, and is expanding its business across all industries including electronics, telecommunications, finance, and automotive. The first concrete commercialization case introduced that day was the ship design document work function in the shipbuilding and marine sector. CEO Kim said, "Previously, it took 2 to 3 months to draft design documents for small ships, but after entrusting this to AI, it is now done within a week. We are researching with the expectation that AI might be able to create the final documents in about 2 to 3 years."

In addition, CEO Kim introduced an ‘Intelligent Maintenance Recommendation Platform’ that recommends repairs for car breakdowns, an ‘AI Assistant’ supporting insurance planners and counselors, an automated solution for legal contract drafting, comparison, and review, a summarization service for books, theses, and patents, and a Q&A chatbot for LG Uplus’s mobile phones and plans sales, attracting attention.