After heated spat with Sam Altman, Elon Musk praises GPT-5

Following a week of public disagreements with Sam Altman, the CEO of OpenAI, Elon Musk has expressed admiration for the company's newly launched GPT-5 model. In a response on X (formerly known as Twitter), Musk referred to one of GPT-5's replies as 'impressive,' signaling a shift in tone regarding the AI technology he has often critiqued. Just before this praise, Musk had asserted that his own model, Grok 4 Heavy, remained a top competitor in the marketplace despite the debut of GPT-5. A user named Kol Tregaskes on X shared a screenshot showing GPT-5's response after it took 34 seconds to address a query. The model candidly stated: 'Short answer: I don't know — and I can't reliably find out,' highlighting a crucial aspect of AI communication. One of the longstanding issues with large language models (LLMs) has been their propensity to generate confident yet incorrect responses, a phenomenon referred to as ‘hallucinations.’ GPT-5's acknowledgment of its limits fosters greater trust among users, akin to how humans often value experts who admit when they lack answers. OpenAI has touted a marked improvement in reducing these hallucinations with the release of GPT-5. The system is comprised of two models: a standard model and a specialized GPT-5 Thinking model, with an automatic router determining which model handles specific queries. Despite these advancements, the new model still fabricates information approximately 10% of the time. Nick Turley, head of ChatGPT, has reiterated caution regarding the model's reliability. In a recent interview with The Verge, he discussed the persistent issue of hallucinations within GPT-5. 'There’s a strong discontinuity between very reliable and 100 percent reliable,' he remarked. Turley emphasized the importance of verifying information, suggesting that users should treat ChatGPT as a supplementary source rather than the primary one for factual verification.