درباره این اپیزود
700,000 people die from suicide every year. Over a billion suffer from mental health conditions. Yet psychiatry still diagnoses depression by asking patients if they feel sad — and current medications work for only 30% of people. Mariam Khayretdinova, CEO of Brainify AI, shares the story of Jay, a brilliant, successful woman who was assessed as “low suicide risk” and died two hours after being discharged. Her keynote makes the case for why AI, trained on massive brain datasets, may be psychiatry’s best hope.
WHAT THIS PANEL COVERS
- Why psychiatry is the only major medical field with zero biological measurement — diagnoses are built entirely on symptomatic questionnaires, leading to 70% misdiagnosis rates
- How Brainify AI’s foundational brain model, trained on data from 500,000 patients, can identify biological patterns invisible to human clinicians and stratify patients by treatment response
- Why “heterogeneity” — where identical symptoms hide vastly different biological causes — is the core challenge, and why depression likely comprises thousands of distinct disorders
- How behavioral data from social media and online activity could predict suicidal individuals and route alerts to emergency systems using data that already exists
- Why building AI for mental health requires 13-trillion-token scale datasets, and why the brain’s diversity across ethnicity, culture, and life experience makes this a fundamentally global challenge
SPEAKER
• Mariam Khayretdinova — Co-Founder & CEO, Brainify AI; Author, “Decoding Despair: How AI is Reshaping Psychiatry”
unDavos is a community-driven summit running during WEF week in Davos, democratizing the conversation around global challenges.
Tags: mental health, AI psychiatry, depression, suicide prevention, brain health, Brainify AI, precision psychiatry, biomarkers, EEG, drug development, mental health stigma, heterogeneity, personalized medicine, Decoding Despair, neuroscience, brain data, unDavos, Davos 2026, WEF
جستجوی اپیزودهای گذشته
قسمت های دیگر در این پادکست
