The partnership between Microsoft and OpenAI has proven to be mutually beneficial. Microsoft’s investment in OpenAI has propelled it to the forefront of AI innovation, expanding the market for its enterprise cloud and AI offerings.
In turn, OpenAI has leveraged Microsoft’s financial resources, computational power, and market reach to transform from a research lab into an AI company that could be valued at around $86 billion. However, Microsoft is taking strategic steps to reduce its dependence on OpenAI, ensuring it captures as much value as possible in the rapidly evolving market for language models and other generative AI technologies.
While OpenAI has made the term GPT synonymous with its brand, Microsoft has been diligently working to associate the term “Copilot” with its own brand. Microsoft’s Ignite conference featured numerous Copilot announcements, including major updates to Copilot for Microsoft 365, the release of Copilots for Service and Sales, and even the renaming of Bing Chat to Copilot.
Microsoft also unveiled Copilot Studio, a low-code tool for customizing Copilot for Microsoft 365 and building standalone copilots. Copilot Studio offers a range of conversational capabilities, including custom GPTs and generative AI plugins. Users can create, test, and publish standalone copilots and custom GPTs, connecting them to external data sources such as SAP. Copilot Studio shares similarities with OpenAI’s GPT Builder and GPT Store with added enterprise features such as data access controls and user controls.
The expansion of large language models (LLMs) is reminiscent of the early days of mobile computing when Apple trademarked the phrase “There’s an app for that.” Today, LLMs are becoming so pervasive that they seem to find their way into every application—sometimes even without a justified use case. The question is, will the new mantra be “There is a GPT for that” or “There is a Copilot for that”?
OpenAI set a precedent in the market for LLMs by keeping its models closed-source, creating a billion-dollar business for its models through its paid API platform and ChatGPT application. However, a growing rival market for open-source models such as Llama, Mistral, Falcon, Cerebras GPT, and MPT is emerging. While setting up an open-source model may not be as convenient as using an out-of-the-box model like GPT-4, it offers benefits like better control over data, performance, infrastructure, and customization.
We’re seeing rapid advances and improvements in tools for training, customizing, quantizing, integrating, and deploying open-source models at scale. Many companies provide services that enable organizations to easily run and fine-tune their custom LLMs on-premise or cloud servers of choice. In fact, the market for open-source LLMs could potentially become much larger than closed-source models like ChatGPT due to its flexibility and numerous corner cases.
Despite its significant stake in OpenAI, Microsoft is not overlooking the market for open-source models. Earlier this year, Microsoft added support for open-source models such as Llama 2, Mistral, and Falcon on its Azure AI Studio. At Ignite 2023, it went a step further, adding “model as a service” support for Llama 2 and Mistral models, making them as easy to use as API calls to OpenAI models.
Google was not the first search engine, Facebook was not the first social network, and Apple was not the first smartphone maker. Similarly, while OpenAI is currently the leading brand in language models and generative AI, it has not yet cemented its place as the dominant player in the field. And another startup might end up usurping its position.
As the market continues to mature and develop, Microsoft is making sure not to put all its eggs in one basket. By embracing both proprietary and open-source models and expanding its Copilot brand, Microsoft is ensuring it remains a key player in the AI market, regardless of how the relationship with OpenAI evolves.
(Copyright: VentureBeat Behind the scenes: How Cvent built an entire webinar using ChatGPT (venturebeat.com)
Chinese
Japanese
Korean
