[Alpha Biz=(Chicago) Reporter Kim Jisun] KT launches super-large AI (Artificial Intelligence) 'Mi: DM' and is targeting the B2B (Business-to-Business) market. The difference KT has put forward is 'AI Full Stack'.

KT held a press conference at the KT Research and Development Center in Seocho-gu, Seoul on the 31st and announced that it officially launched the super-large AI faith.

There are four types of super-large AI models released by KT, including △Basic △ Standard △ Premium △ Expert. The parameters (the parameters) have been subdivided from hundreds of millions of lightweight model basic to hundreds of billions of ultra-large model extremes, allowing companies to use them for their respective purposes. In general, the higher the number of parameters, the better the AI performance.

Currently, the super-large AI ecosystem is led by global big tech companies such as Google, Amazon, and Microsoft (MS). In Korea, Naver announced its entry into the B2B market by introducing HyperClover X. In addition, tech companies such as LG Group's Exaon, SK Telecom's AI Alliance and Kakao are entering the super-large AI market.

KT's strength with the release of Mi:DM is 'AI Full Stack'. AI full stack refers to an integrated product that covers everything from AI infrastructure such as AI semiconductors and clouds to AI application services that customers will use. KT's AI full stack includes AI infrastructure software company day after tomorrow, AI semiconductor company Livelion, subsidiary KT Cloud, FinTech company 2 Digit, AI Startup Upstage and EduTech company Quanta. Cooperation with them has the advantage of reducing costs as well as creating business synergy in each area of expertise.

KT, along with KT Cloud through AI full stack, provides AI cloud farm exclusively for companies of faith as a package. Cloud Farm is a service that allows customers to build and manage their own cloud server farm (Farm) internally to meet the needs of institutions and companies.

In addition, KT Cloud's HAC (Hyperscale AI Computing) service, which can reduce graphics processing unit (GPU) learning costs by about 27%, and Livelion's neural network processing unit (NPU) infrastructure, which reduces inference cost by 50%, will be provided as a customized integrated package for efficient AI development environment.

 

[ⓒ AlphaBIZ. 무단전재-재배포 금지]