The mind-blowing growth of artificial intelligence poses many questions that have no answers yet, the United Nations admitted Thursday at its AI summit, attended by some exceptionally life-like humanoid robots.

The UN is aware that AI technology is racing ahead of the capacity to set its boundaries and directions, and so it brought together some of the best minds on the topic -- whether human or man-made.

The "AI for Good Global Summit", in Geneva on Thursday and Friday is being convened by the UN's ITU tech agency -- and many unaware attendees were startled by the humanoid robots suddenly turning to look at them as they passed by.

"When generative AI shocked the world just a few months ago, we had never seen anything like it. Nothing even close to it. Even the biggest names in tech found the experience mind-blowing," ITU chief Doreen Bogdan-Martin told the summit.

"And just like that, the possibility that this form of intelligence could get smarter than us got so much closer than we ever thought -- including those behind the technology."

The summit is bringing together around 3,000 experts from companies like Microsoft and Amazon as well as from universities and international organisations to try to sculpt frameworks for ensuring AI is used for positive purposes.

Bogdan-Martin painted an alternative nightmare scenario in which AI puts millions of jobs at risk, disinformation spreads widely, and unchecked AI advances lead to "social unrest, geopolitical instability and economic disparity on a scale we've never seen before".

"Many of our questions that we have on AI have no answers yet. Should we hit pause on giant AI experiments? Will we control AI more than it controls us? And will AI help humanity, or destroy it?" she asked.

The robots gathered in Geneva came in many forms: dogs, farm machinery, but also exceptionally realistic avatars, singers, artists and nursing home workers.

With cameras inside their eyes, many were actively following what was going on around them: tracking movement, answering questions, smiling, frowning and even eye-rolling.

The Jam Galaxy Band features humanoid robot Desdemona -- Desi to her friends -- on lead vocals.

Created by roboticist David Hanson, she throws out jazzy lyrics on all sorts of subjects -- love, credit cards, meetings in gardens -- and the band interacts and goes with it.

"It's pretty amazing. You would think it's weird but it's really cool because her AI-generated lyrics are really out there," said soprano saxophone player Dianne Krouse.

"I'm just improvising around that and doing interpretive saxophoning to what she's singing."

Nadine, a robot modelled on University of Geneva professor and virtual human pioneer Nadia Thalmann, was first built in 2013 and can answer questions on the spot.

"I was created to be a humanoid social robot, with human-like physical appearance, to interact with people and to explore the potential of AI technologies," Nadine told AFP.

"I am feeling excited and curious about the AI for Good Global Summit and the potential of AI technologies," the robot said.

"The most interesting person I have met at the AI for Good Global Summit is Professor Nadia Thalmann. She is the one who created me and is a pioneer in the field of AI robotics."

"She's very loyal!", Thalmann added, laughing.

Nadine has improved dramatically over the years and Thalmann said that in the future, "she will be more able to understand her surroundings, ask direct questions -- so not just answer questions -- analyse more, and be more conscious of what is going on.

"Because most robots are not conscious. They can speak but they are not aware of what they say.

"It will take years, because it's quite complex. Self-awareness would be even more complex."