| Mobile Web

Search results for LAB

AI & Enterprise

On-device AI is ready to shake up an AI market dominated by frontier models

AI expert Sigil Wen (시질 웬), who developed the on-device AI model Underdog, said AI models are moving out of data centres and into personal devices. He said most AI tokens will soon be generated on smartphones and laptops. He cited benchmark results showing a 27-billion-parameter Qwen 3.8 27B model on Macs and iPhones scoring higher than Opus 4.6 Max. Wen said large models are unnecessary for daily tasks and stressed speed and security as key advantages of on-device AI.

AI & Enterprise

MSIT launches \'Everyone\'s AI Lab\' picks 12 teams to start service development

South Korea\'s Ministry of Science and ICT and the National Information Society Agency selected 12 teams to advance to the finals of the \'Everyone\'s AI Lab AI service competition\' and held a launch event in Seoul. The contest is part of the \'Everyone\'s AI Growth Ladder Project\' launched in August. A total of 390 proposals were submitted, narrowed to 50 and then 12 through regional presentations. Finalists receive mentoring and resources, with winners chosen in December.

AI & Enterprise

Sam Altman says OpenAI will not go public until it can ensure model safety

OpenAI will not pursue an initial public offering until it is confident about model safety, CEO Sam Altman (샘 알트먼) said. He said there is no specific timetable and that the company must be able to make stronger commitments on safety as AI capabilities expand. Altman said going public could conflict with a safety-first approach and create pressure from Wall Street. He added that an IPO should not be pushed during a period of major change in model power and safety requirements.