Microsoft has expanded its AI initiatives with the introduction of Phi 3, a new language model. Microsoft emphasizes that smaller language models excel at simpler tasks, are user-friendly for organizations with restricted resources, and are easily customizable. They offer cost-efficiency and are compatible with smartphones.
Sonali Yadav, principal product manager for Generative AI at Microsoft said “What we’re going to start to see is not a shift from large to small, but a shift from a singular category of models to a portfolio of models where customers get the ability to make a decision on what is the best model for their scenario,”.
Microsoft suggests that selecting the appropriate language model hinges on the organization’s unique requirements, task complexity, and resource availability. For organizations prioritizing local device applications and tasks with limited reasoning needs or requiring swift responses, small language models like Phi 3 are recommended. This approach avoids the higher costs associated with running LLMs on the cloud, as per Microsoft.
Microsoft highlights that models like Phi 3, capable of offline operation, enable broader AI utilization beyond the capabilities of LLMs. As an illustration, Microsoft mentions a scenario where a farmer identifies crop issues like disease. Leveraging a SLM with visual capabilities, the farmer can capture a photo of the affected crop and promptly receive guidance on pest or disease management.
It’s interesting that Microsoft introduced Phi 3 before the rumored launch of a small LLM by Apple, aimed at enabling AI tasks directly on iPhones.
Source: Money Control