Humanize AI before it dehumanizes us, with Dr. Rana el Kaliouby at SXSW

What It Covers

AI scientist and Affectiva founder Dr. Rana el Kaliouby joins Bob Safian live at SXSW to examine AI's missing emotional intelligence layer, debunk five major AI myths, and outline concrete steps individuals and organizations can take to keep AI development human-centered.

Key Questions Answered

• •AI's EQ Gap: Current AI systems process only the 7% of human communication that is verbal, ignoring the 93% conveyed through facial expressions, vocal tone, and body language. Closing this gap requires building new benchmarks specifically measuring emotional and social intelligence, not just cognitive IQ performance metrics that dominate today's evaluation frameworks.
• •AI Bubble Assessment: Frothy valuations on pre-product, pre-revenue companies raising hundreds of millions at billion-dollar valuations signal localized bubble risk, particularly in circular investment loops between hyperscalers and chip suppliers. However, application-layer companies solving real industry problems represent genuinely early-stage economic opportunity that the market may still be undervaluing.
• •Defensibility Framework for AI Startups: When evaluating AI founders, probe defensibility not just at launch but across the next one-to-three years, since a new model release from Anthropic or Gemini can render a product obsolete overnight. Founders who cannot articulate a durable moat beyond current model capabilities are likely riding hype rather than building lasting value.
• •Human Skills to Prioritize Now: In an AI-driven workplace, collaboration, original communication, critical thinking, and creativity become premium differentiators because AI cannot replicate lived experience or genuine human perspective. Organizations should actively encourage employees to experiment with AI tools, redefine junior roles around human-AI collaboration, and redesign workflows rather than preserving pre-AI processes.
• •AI Safety Guardrails Are Measurable: Every AI model deployment should be tested against standardized safety benchmarks before release, particularly for emotional and therapeutic use cases where vulnerable users interact at 2 a.m. without human oversight. Investors and buyers should require founders and vendors to demonstrate they have explicitly tested models against defined safety criteria, not just functionality metrics.