This is discussed in a new study published in the journal Nature Human Behaviour, writes Axios
Artificial intelligence has the potential to generate numerous creative ideas, but often these ideas turn out to be too similar to each other. This is confirmed by the results of a new study published in the journal Nature Human Behaviour, as reported by Axios.
Researchers from the Wharton School of Business found that 94% of the ideas generated by ChatGPT were repetitive. In the experiment, participants were asked to come up with a toy using a brick and a fan. The most original ideas came from those who relied on their own thinking and internet searches.
Even the new GPT-4 version shows greater diversity compared to the previous GPT-3.5 version, but it still significantly lags behind human creativity. A striking example is that nine ChatGPT users independently came up with their concept called Build-a-Breeze Castle.
“ChatGPT can be useful at the idea generation stage, but only when combined with human imagination and critical thinking. They recommend using specific prompts that encourage AI to create more diverse options.”
The study’s authors emphasize the importance of teaching young people to use generative artificial intelligence not as a replacement, but as a tool to develop their own creativity.
See also:
