When Microsoft announced 365 Copilot in March 2023, it emphasized that OpenAI provided a great advantage by using the GPT-4 model. However, the company’s executives want to reduce 365 Copilot’s dependence on OpenAI due to cost and speed concerns. For this reason, Microsoft is both training its own small models and trying to customize other open source models.
Microsoft CEO Satya Nadella and other leaders are following these efforts closely. Other business units of the company have similarly moved away from OpenAI models and adopted different approaches. For example, Microsoft-owned GitHub introduced alternatives to OpenAI’s GPT-4 model by adding models from Anthropic and Google in October.
Microsoft is customizing various models to make 365 Copilot faster and more efficient. The use of these models aims to reduce the cost of operating the 365 Copilot and pass these savings on to the end user.
Microsoft continues to work on frontier models together with OpenAI. However, the company is taking new approaches to provide 365 Copilot with greater flexibility in terms of cost and speed. These initiatives are expected to make Microsoft’s AI-based products more competitive.
*The visual of the news was provided by Reuters