ChatGPT voice mode cannot time a race, Sam Altman admits

A TikTok creator asked ChatGPT’s voice mode to time a race, only to find that it cannot keep track of elapsed time. Responding to the video, Sam Altman acknowledged that this is a known issue which is unlikely to be fixed for another year.

ChatGPT can examine document pages, create highly convincing images and write computer code for new software. Yet its voice version could be outperformed by an eight-year-old at an extremely simple task: timing something. The TikTok account “huskistaken” demonstrated this in a video, asking the AI to time a race. Although the race lasted just a few seconds, if not less, the AI claimed it had taken almost 10 minutes.

@huskistaken

I swear I was faster

♬ original sound – Husk

ChatGPT voice mode fails to time a race

The video, which received more than 700,000 likes on TikTok, was shown to OpenAI chief Sam Altman during an interview with Mostly Human. Looking slightly embarrassed, he explained that OpenAI’s engineers were already aware of the issue. The problem is reportedly that the voice model featured in the TikTok video does not have the tools required to start a stopwatch. It is not expected to be resolved for another year.

The video also highlights another issue: rather than admitting that it could not time the race, the AI model chose to make up a number.

The voice mode still has many shortcomings

While ChatGPT’s voice mode delivers remarkably realistic and smooth interactions, it still has plenty of weaknesses. The TikTok account behind the video shown to Sam Altman has, in fact, amassed millions of views through clips revealing this AI’s mistakes. In one video, which reached 1.2 million views, ChatGPT’s voice mode tells the user that the word “December” contains the letter X.

@huskistaken

Learned something new!

♬ original sound – Husk

OpenAI intends to make its voice AI more intelligent

Fortunately, OpenAI is expected to improve its voice AI in order to reduce the number of similar errors. During the Mostly Human interview, Sam Altman said that OpenAI would add more intelligence to its voice model.