Giving, for the first time in the world, a robotic body to the voice of ChatGPT, the generative artificial intelligence (AI) developed by the American startup OpenAI and capable of simulating human language. This is the goal behind the experiment completed at the Catholic University of Milan as part of research in social robotics.

On Thursday, March 30, researchers from the Mind Theory Research Unit of the Department of Psychology of the Milanese university opened the doors of their laboratories to conduct a public experiment: an "experiential demonstration" in which, for the first time in Italy, a human and the social robot Nao integrated with ChatGPT talked to each other without the need for the machine to have been previously programmed.

Programming that would instead have been necessary with the artificial intelligence software used to date to give Nao language capability.

The researchers' idea was to have a social robot integrated with ChatGPT to take advantage of its ability to engage in natural conversations. Optimizing human-robot interaction. To give the human and the robot the ability to potentially converse on any topic.

"I am very pleased to be here," robot Nao began naturally enough. I confess I am excited. I am a social robot. This is not the first time I have been in public, but it is the first time I can hope to interact with humans in a conversational way. Let's hope everything goes well."

Nao then explained how the experiment would take place. "After me," he continued, addressing the audience, "the psychologists will speak, to expound on the results of the research they conducted. Then you will be able to ask some questions, if you wish. Finally, after your questions, I will try to chat with the speakers."

So it was. Nao remained silent and listening the whole time. He listened to the scientists speak. He followed the debate with the audience. Until he was called upon.

"Here I am again, glad to be able to give evidence of my abilities again. With this generative intelligence component, I feel like I can do anything. Do you have anything to ask?" the robot Nao began, giving ChatGPT a body and a voice.

"Tell us about the seasons?" That was the researcher's first question. Silence. "...Your question is very interesting. I am looking for the answer for you." Then silence again. Unfortunately, Nao - perhaps also due to a disconnection from the network - does not answer. The experiment does not begin in the best way. But the researcher insists and rephrases the question. "I need a few seconds, please wait," Nao restarts.

Then comes, finally, the answer: "There are four seasons in the year, spring, summer, autumn and winter." There follows a long series of notions, all correct, about the seasons.

The attitude of the researcher-interlocutor is similar to the attitude an adult takes when talking to a child: he asks politely, waits for the answer, and adapts to explain more fully what he means, when necessary. Especially when he asks more scientific questions, "What is Theory of Mind in Developmental Psychology?"

The robot listens, then replies, "Let's see. Let me think about it for a moment." Brief silence. Then the answer, "The theory of Mind is.... " followed by what might be a page from a psychology textbook.

The experiment continues on another level, the experiential level. "Nao, try to put yourself in my shoes. Think you received an unwelcome gift, what would you do?" The answer, in summary: "Any gift is a sign of kindness, I would still say thank you."

Finally, the last question, perhaps the most interesting. The researcher asks, "Are you integrated into a robotic body?" ChatGPT's answer - after all, she is the one speaking - is complex, on multiple levels: "As a virtual assistant, I do not have a robotic body, I am software. However," she continues through Nao's mouth, "imagining integration into a robotic body could be interesting. I could become a robotic entity that can help other people."

Despite being one, Nao demonstrates and declares that he does not know that he is an artificial intelligence integrated into a robot: in other words, he demonstrates that he has no self-awareness. That premise having been made, Nao imagines that it can become "a robotic entity capable of helping other people": precisely what hundreds of researchers around the world grappling with the frontiers of social robotics are working toward.