• AIPressRoom
  • Posts
  • Microsoft is Reportedly Engaged on AI Chips to Prepare LLMs

Microsoft is Reportedly Engaged on AI Chips to Prepare LLMs

Microsoft is reportedly working by itself AI chips that can be utilized to coach LLMs

After investing billions within the agency that created ChatGPT, OpenAI, Microsoft is reportedly working by itself AI chips that can be utilized to coach giant language fashions or LLMs. Microsoft is making waves within the synthetic intelligence (AI) house.

Based on The Info, which cited two sources with firsthand information of the undertaking, the software program behemoth has been engaged on the microprocessor since as early as 2019. A restricted variety of Microsoft and OpenAI personnel who’re evaluating the expertise reportedly have already got entry to it underneath the code title Athena.

Microsoft hopes that the chip would outperform these it now purchases from different suppliers, saving it money and time on its costly AI initiatives. Based on the examine, different well-known tech companies like Amazon, Google, and Fb additionally produce their very own AI processors.

For the AI analysis agency OpenAI, Microsoft has already constructed a supercomputer that may practice huge numbers of fashions. To assist ChatGPT and the Bing AI chatbot, the enterprise makes use of tens of hundreds of Nvidia A100 graphics chips linked to a community for the supercomputer. A “large, cutting-edge supercomputer” was going to be constructed, thus it invested US$1 billion in OpenAI in 2019.

Microsoft created this supercomputer to have the processing capability crucial to coach and retrain an increasing variety of AI fashions over prolonged intervals utilizing large quantities of information.

Nidhi Chappell, Microsoft’s head of product for Azure high-performance computing and AI, said, “One of many issues we had discovered from analysis is that the bigger the mannequin, the extra knowledge you’ve got, and the longer you may practice, the higher the accuracy of the mannequin is.

Google’s TPU AI Chip

Google revealed that it has created an AI chip dubbed the Tensor Processing Unit (TPU) that was made solely for machine studying actions final 12 months. Based on claims, the TPU makes use of little energy and might do billions of processes per second.

TensorFlow, Google’s open-source machine studying software program framework, is meant for use with the tensor processing unit.