Why is Microsoft drifting away from OpenAI?

2023-08-15

It seems that Microsoft's honeymoon period with OpenAI is coming to an end. Microsoft appears to be slowly distancing itself from OpenAI, prioritizing its Azure cloud and taking measures to protect its own interests. With a focus on becoming an enterprise-friendly entity, Microsoft's growth largely depends on maintaining customer trust.

Recently, OpenAI has not been performing well, putting customers' data at risk with ChatGPT. Large tech companies like Apple, Spotify, Wells Fargo, Samsung, JPMorgan Chase, and Verizon have long abandoned ChatGPT and prohibited its use by their employees.

Companies like Apple have legitimate concerns about their employees inadvertently sharing sensitive project details through ChatGPT. These data could be accessed by OpenAI's reviewers. Research has shown that certain language models can extract training data through their chat interfaces. Apple recently developed its own internal chatbot, referred to by employees as "Apple GPT."

Recently, OpenAI publicly launched GPTbot, an automated web crawler used to collect publicly accessible data for training AI models. However, this has not helped their situation. Microsoft relies on OpenAI's base models to provide services to its customers, and thus, it has been affected.

Microsoft knows that enterprises and their employees need ChatGPT because it has practical value in enhancing coding and generating creativity. Employees find themselves in a dilemma - whether to use ChatGPT or not. Some still secretly use it even if their organizations prohibit its use.

Recently, Reuters released a report revealing that an increasing number of US workers rely on ChatGPT to complete basic tasks, despite some concerns, leading companies like Microsoft and Google to restrict its use.

To overcome trust issues between enterprises, Microsoft recently announced Azure ChatGPT, marketed as an enterprise version of ChatGPT where they don't have to worry about their data. Surprisingly, Microsoft acknowledged the drawbacks of OpenAI's ChatGPT when making the announcement.

"ChatGPT has the potential to expose confidential intellectual property. One option is to block enterprise access to ChatGPT, but people will always find workarounds." Microsoft said, "Azure ChatGPT is our enterprise edition option. This solution provides a similar user experience as ChatGPT but as your private ChatGPT."

Microsoft also explicitly states that customer data is "completely isolated from OpenAI-operated data." Did OpenAI fail to protect data and use it to train GPT-4? When it released the ChatGPT API in March, the company stated that OpenAI retains API data for 30 days but no longer uses data sent via the API to improve its models.

When enterprises make API calls using GPT-4, OpenAI wants to retain user history for further training. In Azure ChatGPT, Microsoft places OpenAI's model behind a user's Azure account-specific endpoint. Enterprises have been interested in GPT-4, hence the demand for private endpoints. Amazon is also doing the same, like Bedrock.

Azure is the only hope

Currently, Microsoft is gaining momentum by launching a range of generative AI projects through Azure in collaboration with OpenAI, with Azure's revenue growing by 26%. Now, Microsoft doesn't want the Azure brand to suffer due to OpenAI's privacy issues.

When Microsoft partnered with OpenAI, it deployed OpenAI's technology through APIs and Azure OpenAI services, enabling enterprises and developers to build applications based on GPT, DALL·E, and Codex. They also worked together to incorporate OpenAI's technology into applications like GitHub Copilot and Microsoft Designer.

However, a common question among customers is whether data submitted to Azure OpenAI services remains in Microsoft Azure or is passed on to OpenAI at any time. Microsoft claims that data submitted to Azure OpenAI services is retained in Microsoft Azure and not passed on to OpenAI for model predictions, with Azure having complete control and management over the data and OpenAI.

Microsoft holds enough shares in OpenAI that their ultimate goal is to integrate features like GPT into Azure and Office365, providing services to enterprise customers, which is something we might see happening. One user said that OpenAI targets private consumers while Microsoft focuses on enterprises.

He also gave an example from his own company. "I can use my own organization as an example. We are an investment bank in the EU that deals with green energy. If it were legal, we would definitely use GPT, but it's not legal, and considering that part of their financial model is to steal as much data as possible, it may never be legal."

Microsoft is concerned about the negative perception of OpenAI's data security among enterprises. By launching Azure ChatGPT, they are working to rebuild trust between enterprises and attract more customers. They don't want to be associated with OpenAI's name on Azure, which is why they put the Azure name at the forefront.

Azure ChatGPT is just part of Microsoft's overall strategy to achieve comprehensive dominance in the enterprise IT field. Additionally, this is not the first time Microsoft has had disagreements with OpenAI. Earlier, The Wall Street Journal reported complaints from within Microsoft about reduced spending on its internal AI and OpenAI's restriction on most Microsoft employees accessing its technology's internal workings.

RECENT AI NEWS

RECENT AI TOOLS