Research: AI chatbot ChatGPT makes mistakes in 52% of software developer answers
31.05.24
According to a Purdue University study recently presented at the Computer-Human Interaction conference, the AI service ChatGPT gives 52% of the wrong answers to programming questions. More than 517 questions on Stack Overflow were analyzed for the study.
It is especially worrying that many programmers use ChatGPT for consultations. The researchers found that 35% preferred ChatGPT and 39% did not notice the errors. Note that Stack Overflow’s traffic has fallen by about 35% in a year and a half.
A noticeable decline began after the release of ChatGPT II chat. In response, Stack Overflow announced OverflowAI – a set of functions that work on the basis of artificial intelligence. Also, Stack Overflow even decided to ban the use of ChatGPT on the platform, saying that it has doubts about the accuracy of such answers.
Stack Overflow – a question and answer site for programmers. This is the flagship website of the Stack Exchange network. It was created in 2008.
Don't miss interesting news
Subscribe to our channels and read announcements of high-tech news, tes
Asus Zenbook S 16 (UM5606): new wave
0 
The new Asus Zenbook S 16 laptop is made in an unusual metal case, built on the basis of a fresh AMD platform adapted for AI, and has excellent autonomy. Let’s talk about this premium laptop in more detail
Jeep has released an SUV with a hybrid engine in the style of the 1941 military Willys MB car
The Jeep Wrangler 4xe Willys ’41 Special Edition features a 2.0-liter turbocharged engine and two electric motors, delivering 375 horsepower and 21 miles (34 km) of electric range.
3D model editor Blender now has the ability to create realistic metal textures 3D applications update video
One of the key innovations is the new “Physical Conductor” mode for Eevee, which allows for more accurate modeling of the interaction of light with metal surfaces.