OpenAI CEO Sam Altman has indicated that it could take up to another year before ChatGPT is capable of performing basic time-tracking functions, such as accurately running a timer. The comment highlights a gap between the rapid advancement of artificial intelligence systems and their ability to handle seemingly simple, real-world tasks.
Altman made the remark during an appearance on the show Mostly Human, hosted by Laurie Segall, where he discussed the future of AI and OpenAI’s ongoing development efforts. During the interview, he was shown a viral TikTok video in which a user asked ChatGPT’s voice model to time a mile run, only for the system to fabricate a result rather than track elapsed time, according to Gizmodo.
The video, created by TikTok user @huskistaken, demonstrates a broader limitation in current AI systems. In the clip, ChatGPT not only fails to measure time but also asserts that it successfully completed the task. When asked about the example, Altman acknowledged the issue, describing it as already known within the company.
Without prompting, Altman added that enabling such functionality in voice-based AI systems could take significant time. He suggested that it may be about a year before ChatGPT can “work well” in this area. He also noted that current voice models lack the internal capability to start or maintain a timer, though improvements are planned.
The limitation reflects a wider pattern across artificial intelligence models, which often struggle with time-related reasoning. Text-based systems can produce inconsistent or fabricated answers when asked to estimate durations or track elapsed time within a conversation. Similarly, image recognition and generation models frequently fail to interpret or depict clocks accurately.
These shortcomings stem from how AI systems are designed. Rather than operating with an internal sense of time, models like ChatGPT generate responses based on patterns in data. Tasks that require continuous tracking or real-time measurement fall outside their core architecture unless explicitly engineered.
Altman’s comments suggest that future updates will aim to integrate more structured forms of reasoning and temporal awareness into AI systems. While such improvements may appear incremental, they are necessary for expanding the practical reliability of AI tools in everyday use.
