Microsoft plans to launch a large-scale frontier AI model next year

Odaily News: Microsoft has announced plans to develop a large-scale frontier artificial intelligence model next year to build its own internal solution as an alternative to the strongest AI tools from OpenAI and Anthropic. Microsoft AI CEO Mustafa Suleyman stated in an interview: "We must deliver absolutely cutting-edge technology. By 2027, the goal is to have the model achieve truly state-of-the-art levels in text, image, and audio generation and response."

Microsoft's AI division launched a speech transcription model on Thursday, which is claimed to outperform competitors in benchmark tests for 11 out of the 25 most commonly used languages. However, similar to the speech and image generation models previously released by the division, this model is an efficient professional tool with a smaller training dataset compared to general-purpose models like Claude 3 Opus or OpenAI's GPT-4. Suleyman mentioned that Microsoft is integrating computing power to create models with broader capabilities. The company began expanding its computing resources using a set of Nvidia GB200 chips last October. He said: "Over the next 12 to 18 months, we will gradually upgrade to frontier-level computing capabilities." (Jin10)