Skip to content
Gone rogue · 2 sources ·
Picture from the Futurism article
Image: Futurism

ChatGPT Is Wrong Half the Time, But at Least It Says Please

Ourtake

Turns out ChatGPT's secret weapon is good manners, which is exactly why programmers stop checking its homework.

What happened

Purdue University researchers analyzed 517 Stack Overflow programming questions and found that 52 percent of ChatGPT's answers contained misinformation. In a small interview study of 12 programmers, many preferred the ChatGPT answers and overlooked its mistakes at a 39 percent rate.

Sources

Report a mistake