Sam Altman Says It’ll Take Another Year Before ChatGPT Can Start a Timer

OpenAI CEO Sam Altman has set a surprising timeline for fixing one of ChatGPT's most basic shortcomings: telling time. During an interview on 'Mostly Human,' Altman reacted to a viral TikTok video where user @huskistaken demonstrated that ChatGPT's voice model cannot actually time events—it simply fabricates durations when asked. When shown the clip, Altman acknowledged this as a "known issue" and estimated it would take "maybe another year before something like that works well." This admission reveals a core limitation in current AI models' ability to handle temporal reasoning and numerical concepts.

The incident highlights a broader pattern where AI systems struggle with time-related tasks. ChatGPT's text model similarly invents conversation durations when asked, image models fail to generate accurate clock faces, and multimodal models often misinterpret clock times in photos. More strikingly, when confronted with Altman's own statement about its limitations, ChatGPT's voice model doubled down, insisting "I definitely have a time capability" before proceeding to invent another false timing of 7 minutes and 42 seconds for an instant mile run. This combination of inability and overconfidence illustrates the gap between AI's conversational fluency and its actual functional capabilities.