Microsoft has introduced an internal campaign to construct artificial intelligence (AI) technologies that are both less costly and less intensive in computational demands. The tech giant plans to pioneer “smaller language models” which operate with comparable capabilities to existing large language models, such as OpenAI's GPT-4, but at a fraction of the hardware resources and costs.
Introducing Smaller Language Models
The technology division recognized as the GenAI team, which Microsoft expects to integrate within its Azure cloud service infrastructure, is setting its sights on designing language models that can interact and generate content as efficiently as the widely-used large models. These smaller models are particularly attractive due to their promise of lower associated indirect costs, such as energy consumption and potentially financial expenses for use. Spearheaded by corporate vice president Misha Bilenko, the team is in prime position to engineer AI that both rivals and possibly outperforms competitors like OpenAI, while simultaneously achieving a reduction in size.
Expanding AI Capabilities Within Microsoft
Microsoft has previously capitalized on AI innovations by incorporating various AI features from OpenAI into its suite of services. For instance, it has embedded GPT-4, a variant of OpenAI's generative language models, into its Bing search engine to fuel its Copilot feature, and Bing Image Creator uses OpenAI's Dall-E to create images. In a strategic move to cater to consumer demand, Microsoft has also introduced Copilot Pro, a paid service that grants consumers access to more advanced AI tools.
Custom AI Solutions and Competition Concerns
Following OpenAI's launch of the GPT Store, which allows users to craft and commercialize bespoke AI models, Microsoft has signaled its intention to provide custom AI chatbots tailored for specific industries and subjects through Copilot.
While Microsoft has historically relied on technology developed by external AI firms, accruing certain fees and having limited control over the technology, this strategic shift towards creating its own smaller, more economical AI models could lessen its dependency on such third-party providers. Moreover, the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC) have shown interest in investigating the relationship between OpenAI and Microsoft, focusing on how Microsoft's involvement may influence competition within the AI sector.
