Ad
Short

Did Microsoft choose this slide wisely? Kevin Scott, Microsoft's CTO, just showed a graph at Microsoft's Build developer conference that was probably meant to illustrate the potential (?) exponential growth of AI. Taken at face value, the graph implies that an OpenAI AI model many times more capable than GPT-4 will be released this year. If the graph was created without thinking about future AI models, it begs the question of why Microsoft a) doesn't know better and b) wants to fuel AI hype only to disappoint people later. Soon after, OpenAI CEO Sam Altman took the stage and repeated what he's been saying for months: that AI models will get "smarter" with more training. Altman did not provide any specific information about GPT-5.

Image: Microsoft (screenshot on YouTube)
Ad
Ad
Ad
Ad
Short

Baidu says it has developed technology that combines graphics processors from different manufacturers into a single computing cluster for training AI models, CEO Robin Li explained in a quarterly earnings call. The platform also works efficiently with clusters of hundreds or thousands of GPUs, Li said. Baidu aims to transform itself from an internet company into an AI company by 2025, with its ERNIE generative AI model becoming the new core of its products. Chinese AI companies have been hurt by U.S. sanctions that prevent Nvidia, in particular, from selling its most capable AI training graphics cards in China. If Li's claim proves true, it could be a big step forward for China's AI industry and for addressing the chip shortage in AI training more broadly.

Ad
Ad
Google News