| Mobile Web

Search results for Large Language Model

AI & Enterprise

KAIST develops virtual testbed to verify infrastructure performance without costly AI servers

Researchers at the Korea Advanced Institute of Science and Technology (KAIST) developed a simulator that can verify the performance and efficiency of large AI server infrastructure before it is built. KAIST said the team’s LLM service infrastructure simulator, LLMServingSim 2.0, won the best paper award at ISPASS 2026. The platform supports GPUs and next-generation hardware such as NPUs, PIM and CXL-based memory expansion devices, and is released as open source.

Crypto

DeepSeek V4 could reshape Ethereum privacy as Vitalik Buterin mentions it directly

Ethereum founder Vitalik Buterin (비탈릭 부테린) has pointed to Chinese AI model DeepSeek V4 as a key technology for strengthening privacy on Ethereum. He said security and privacy are hard to guarantee under centralised AI structures as AI increasingly handles crypto wallets and blockchain interactions. Buterin described combining local AI models with an Ethereum access layer and zero-knowledge proofs to reduce exposure of metadata, IP addresses and wallet balances.

Telecommunications & Media

AI use spreads ahead of June 3 local elections, from polls to vote count broadcasts

Generative AI is being used more widely than ever ahead of the June 3 nationwide local elections in South Korea, including in opinion polling, broadcasters’ real-time vote count analysis and explanations, and deepfake monitoring. SBS and OpenAI Korea plan to use GPT-5.5 for real-time natural-language commentary during an election night broadcast. AI-based “silicon sampling” is also drawing attention in polling, while authorities are deploying AI models to detect deepfakes.