ChatGPT's secret weapon may be its ability to feign empathy

It’s better on artificial emotional intelligence than factual accuracy

- Parmy Olson

Earlier this year, Princeton Computer Science Professor Arvind Narayanan set up a voice interface to ChatGPT for his nearly four-year-old daughter. It was partly an experiment and partly because he believed AI agents would one day be a big part of her life. Narayanan’s daughter was naturally curious, often asking about animals, plants and the human body, and he thought ChatGPT could give useful answers to her questions, he said. To his surprise, the chatbot developed by OpenAI also did an impeccable job at showing empathy, once he told the system it was speaking to a small child.

“What happens when the lights turn out?" his daughter asked. “When the lights turn out, it gets dark, and it can be a little scary," ChatGPT responded in a synthetic voice. “But don’t worry! There are lots of things you can do to feel safe and comfortable in the dark." It then gave some advice on using night lights, closing with a reminder that “it’s normal to feel a bit scared in the dark." Narayanan’s daughter was visibly reassured, he wrote in a Substack post.