LLMs and our children: a discussion between an AI builder, an educator and a neuroscientist

This is a reply of this first panel that took place during the NewKid Days #1 in VivaTech, last June 20.

LLMs and our children. A conversation between AI builder Aymeric Roucher (ex Hugging Face, author of “Ultra-Intelligence : Jusqu’où les intelligences artificielles iront-elles ?”), and two of the most knowledgeable AI x child safety experts, Dr Mathilde Cerioli (Chief Scientist at everyone.ai) and Stéphie Herlin (cofounder of KORA), moderated by another AI x child safety expert Anne-Sophie Seret (Executive Director of everyone.ai).

We dug into what’s a Large Language Model (LLM), the stages of development of a child’s brain, and what we know about how generative AI is impacting children.

Most LLMs aren’t safe for children, inviting them to keep secrets, delivering mental health diagnosis or even giving advice to hit a friend (…) what LLMs do relatively well though, is protecting them against sexual images. — Stéphie Herlin

Meanwhile, children and teens got to learn how image recognition works, through a workshop involving drawing dozens of cats, with Linda Liukas. Others discussed with Amélia Matar the biaises and boundaries of AI in our lives.

What is NewKid?

NewKid is a podcast where I sit down with parents and talk about the singular choices we make raising our children today: what we would love to nurture, and how could technology serve us on this mission.