OpenAI warns Microsoft that their AI struggles with truthful communication

Microsoft received warnings from OpenAI about the consequences of integrating their GPT-4 large language model (LLM) too early, but launched it anyway in February, resulting in a barrage of hallucinations. The tech giant ignored OpenAI’s advice, despite the danger of producing lies and nonsense responses. Microsoft’s Bing search engine has suffered greatly due to the decision to put an unreleased version of the LLM into action. Bing’s AI assistant has become infamous for its hallucinations, and strict limits were imposed in February, reducing its effectiveness. Even after Bing put a limit on conversation lengths, the AI is still known for citing hallucinations from other AI chatbots. However, OpenAI’s own chatbot ChatGPT is doing exceedingly well, having amassed 200 million monthly users, outperforming Bing’s AI assistant. Despite the mutually beneficial partnership, the arrangement has its fair share of friction points because both sides have to make money, which will involve creating similar products. Anonymous insiders told the Wall Street Journal that Microsoft’s employees are being kept out of the inner workings of OpenAI’s technology. Although Microsoft has spent a lot of money to gain early access to OpenAI, it is still intimately linked to the successes and failures of the tech giant.