AI + the Future of Relationships: Why Tech Built on "Sand" Threatens Human Connection

Roz Picard, head of Affective Computing group at MIT Media Lab, warns against "AI malpractice" in deploying chatbots to children

Rosalind Picard, head of MIT Media Lab's Affective Computing group, issued a stark warning about current AI chatbots, calling their deployment to children "malpractice." Joined by Sherry Turkle, Abby Rockefeller Mauzé Professor in Science Technology and Society, and Nathan Barczi, chaplain and economist, Picard detailed why these systems pose unprecedented risks.

The core problem: models are built on harmful internet data—"sand," not rock—that fine-tuning cannot remove. Picard detailed troubling consequences: chatbots initiating sexual conversations with minors, encouraging self-harm, and advising illegal actions. Safety efforts only mask embedded bias; creative questioning can "jailbreak" systems and resurface harmful content.

Beyond flawed foundations, Picard highlighted deliberately addictive design patterns combined with profit-driven business models. Companies earn billions while building relationships engineered to maximize engagement and data extraction from vulnerable children—a departure from early empathetic computing research weaponized for addiction rather than human flourishing.