OpenAI, the company behind ChatGPT and GPT-4, reportedly warned Microsoft to go slow with integrating the large language models (LLMs) into Bing search to avoid inaccurate and unpredictable responses. OpenAI was concerned about Microsoft implementing an unreleased version of GPT-4 too early.
Apparently, Microsoft didn’t listen. For a short stint after launching the revamped Bing search, people went wild sharing screenshots of Bing Chat doing all kinds of crazy stuff, like gaslighting people, lying, sulking and more.
Microsoft responded by capping the length of conversations people could have with Bing Chat to limit the potential for the LLM to go haywire. Since then, the company has eased restrictions as updates limited the crazy responses.
The report comes from the Wall Street Journal, which also detailed ongoing tensions between OpenAI and Microsoft. Despite the companies’ partnership, they offer competing products. For example, WSJ report details how Microsoft executives were anxious over the launch of ChatGPT since it was already integrating LLMs into Bing.
Currently, Microsoft licenses OpenAI’s models and technology for use in products like Bing, Azure, Office, Windows and more. Meanwhile, Microsoft is OpenAI’s exclusive cloud partner and the company’s cloud service power all OpenAI workloads, API services and research. And there’s also Microsoft’s multibillion investment in OpenAI, rumoured to be around $10 billion USD (about $13.3 billion CAD).
In a recent interview with Wired, Microsoft CEO Satya Nadella acknowledged the companies’ competition.
“I felt OpenAI was going after the same thing as us,” Nadella told Wired. “So instead of trying to train five different foundational models, I wanted one foundation, making it a basis for a platform effect. So we partnered. They bet on us, we bet on them.”
However, Nadella also dodged questions about whether Microsoft tried to acquire OpenAI.
Image credit: Microsoft