Shulou(Shulou.com)11/24 Report--

Thanks to CTOnews.com netizen OC_Formula for the clue delivery! CTOnews.com, July 20, LG officially released the multimodal large language model EXAONE 2.0 last night, which supports Korean and English and can be used in new materials, new drug development and other fields.

According to reports, EXAONE 2.0 has learned "about 45 million professional documents such as patents and papers obtained through partnerships, as well as 350 million pictures." In order to solve the high cost problem of super-large AI, EXAONE 2.0 not only deals with a variety of information such as large-scale language model (LLM), image and language, but also carries out lightweight design.

In addition, considering that most of the existing professional knowledge data are in English, LG developed EXAONE 2.0 into a bilingual model that can understand and answer both Korean and English at the same time, and the amount of learning data is more than four times that of the existing model. The LG AI Institute explained that while maintaining the same performance as the existing models, the EXAONE 2.0 language model reduced reasoning processing time by 25 per cent and memory usage by 70 per cent, saving about 78 per cent.

CTOnews.com learned from LG's official website that EXAONE 2.0 provides three major service platforms, namely ExaOne Universe, ExaOne Discovery and ExaOne Atelier.

LG calls ExaOne Universe a "professional interactive AI platform" that can effectively help experts with production. For the questions that require professional knowledge, the platform can provide the latest professional data in various fields, find out the basis, and give the answer of inference.

AI researchers, partner universities and others will begin to provide Closed Beta (private testing) services for the ExaOne Universe platform from the 31st of this month, and will begin to provide formal services that are available to all researchers in the company in September.

ExaOne Discovery platform can make use of multimodal AI technology to analyze not only the text of professional literature, but also the non-text information such as molecular structure, chart, table, image and so on. LG AI Institute predicts that through dialogue with AI, professional literature review, material structure design, material synthesis prediction can be carried out, and the time required for research and development will be reduced from 40 months to 5 months.

In the fourth quarter of this year, ExaOne Discovery platform services will be provided to researchers in the chemical and biological fields of the LG Group.

The ExaOne Atelier platform can express "images in language", match 350 million copyrighted images learned by EXAONE 2.0 with text, and provide special functions for image generation and understanding.

ExaOne Atelier platform will provide services for professional designers of LG Group this quarter.

According to Lu Mingxun, dean of the LG AI Research Institute, EXAONE 2.0 currently serves only B2B (between businesses), but it is likely to expand to B2C (between businesses and consumers) in the future.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.