‘AI Hallucinates’: Sam Altman warns about putting blind faith in ChatGPT

Since its introduction to the public in late 2022, ChatGPT has emerged as a dominant AI chatbot, becoming an indispensable tool for countless users. However, Sam Altman, the CEO of OpenAI, has issued a stark warning about the dangers of placing unwavering faith in the technology, citing its tendency to produce inaccurate information or 'hallucinations'. In the inaugural episode of the OpenAI podcast, Altman remarked, "People have a very high degree of trust in ChatGPT, which is intriguing because AI can hallucinate. It should be regarded as technology that demands a healthy skepticism." He emphasized the importance of acknowledging ChatGPT's limitations, stating, "It’s not super reliable… we need to be honest about that." The phenomenon of hallucination in AI chatbots, such as ChatGPT, refers to the generation of confident yet false responses. This issue can stem from various factors, including biased training data, a lack of grounding in factual information, the inherent pressure to provide responses, and the predictive nature of text generation. Currently, no major AI company asserts that its chatbots are immune to these inaccuracies. During the podcast, Altman also reiterated a previous prediction regarding the future of AI: he believes his children will never surpass AI in intelligence. However, he expressed optimism, stating, "But they will grow up vastly more capable than we did, equipped to accomplish things that we cannot even fathom." When questioned about the potential introduction of advertisements within ChatGPT, Altman responded thoughtfully, saying, "I’m not totally against it. There are instances where I find ads beneficial. For example, ads on Instagram can be quite effective—I’ve purchased several items because of them. However, implementing ads in a way that doesn’t disrupt user experience would be challenging. The burden of proof in that regard has to be quite high, ensuring that it feels genuinely useful to users and does not interfere with the chatbot's output." Stay updated with the latest technology news and insights.