The Model Can’t Relate: A poet’s rebellion inside the AI machine with Danielle McClune

What if the people building AI are so caught up in the rush to market that they’ve forgotten to ask the most important question: what does this mean for humanity? In this refreshingly honest episode, we explore the human side of artificial intelligence with Danielle McClune, a writer and poet who has spent the last years at the epicenter of AI development at Microsoft, training conversational models and crafting the prompts that shape how AI communicates with millions of users worldwide.Danielle takes us behind the scenes of AI development with a perspective that’s rare in the tech industry – one grounded in creative writing, poetry, and a deep concern for preserving our humanity in an increasingly automated world. From her Substack “Soft Coded” writing that challenges the industry’s relentless optimism to her daily work training models to sound human while remembering they’re not, Danielle offers a critical yet nuanced view of where AI is headed and what we might be losing along the way.Throughout our conversation, Danielle reveals the absurdity of charging users for saying “please” and “thank you” to AI while encouraging human-like interaction, questions why we’re bolting chat interfaces onto existing software instead of reimagining human-computer interaction, and argues for maintaining the “uncanny valley” as a crucial reminder that we’re not talking to someone with a childhood. Her vision for AI as a public utility and her insights into what the technology might look like if women had led its development offer provocative alternatives to the current Silicon Valley narrative.In this conversation, we explore:Why saying “please” and “thank you” to AI reveals deeper contradictions in how we’re building the technologyThe rush to add chat interfaces to everything instead of reimagining user experiences from scratchWhy the uncanny valley might be a feature, not a bug, in human-AI interactionHow “vibe checks” and human intuition remain essential in evaluating AI outputThe case for treating AI as a public utility rather than private corporate propertyWhy training AI models feels like “raising a toddler” and often becomes “women’s work”This episode is an invitation to slow down, ask harder questions, and remember that behind every AI interaction is a human being whose life might be changed – for better or worse – by the choices we make today.Resources MentionedSoft Coded is Danielle’s excellent SubstackRuined by Design – Mike Monteiro’s bookDesign for the Real World – Victor PapanekConnect with DanielleLinkedin: https://www.linkedin.com/in/danielle-mcclune-2b35b95b/Substack: https://softcoded.substack.com/BioDanielle McClune is a writer and poet embedded in the frontier of AI development at Microsoft, where she has spent the last two years training conversational models and crafting the prompts that shape how AI communicates with users worldwide. Originally from Wisconsin with a bachelor’s degree in creative writing and dreams of becoming poet laureate, Danielle found her way to Seattle and the tech world through UX writing. She later earned her MFA in Arts Leadership from Seattle University, focusing on arts nonprofits and public policy. Through her Substack writing, Danielle offers a rare critical perspective on AI development from insideSend us Fan MailFollow us on Instagram: https://www.instagram.com/poetsandthinkerspodcast/Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/poets-thinkers/id1799627484Subscribe on Spotify: https://open.spotify.com/show/4N4jnNEJraemvlHIyUZdww?si=2195345fa6d249fdSend your ideas, feedback and guest recommendations to [email protected]