Search
Following the scaling back of its partnership with OpenAI, Microsoft is considering incorporating a broader range of cost-efficient technologies within the generative AI space. Citing sources from Axios, Microsoft is currently assessing the viability of utilizing its self-developed and adapted Chinese open-source model, DeepSeek, within its AI assistant, Copilot. This move comes in response to escalating costs attributed to the "tokenmaxxing" trend—a practice where AI models generate excessively lengthy outputs to maximize token usage, driving up computational expenses. By hosting the DeepSeek model in-house, Microsoft aims to gain greater control over operational costs. The company is anticipated to roll out a more affordable version of Copilot in the upcoming weeks, along with disclosing the specific AI model employed.
-
C114 Communication Network -
Communication Home
