OpenAI's CEO says the company's large language model has been misunderstood and that he has shifted his stance on AGI.
OpenAI’s August launch of its GPT-5 model turned into a public stumble. The livestream showed glitches, with the system producing charts that featured clearly incorrect figures. In a Reddit AMA with OpenAI employees, users complained that the new model didn’t feel friendly and urged the company to restore the previous release. Above all, critics argued GPT-5 failed to meet the stratospheric expectations OpenAI had been building for years. Billed as a major advance, GPT-5 may improve some functions, yet it still operated within the same framework. The stumble and fallout prompted internal reflection at the company about product messaging and the path forward for advanced AI. The CEO acknowledged the pushback and signaled a shift in how he assesses AGI development moving forward.