Dr. Rana el Kaliouby opened her plenary at the Rhodes Forum on Technology and Society with a show of hands. Would you let AI deliver your morning news? Write your code? Hire, promote or fire your colleagues? Step in when your child is being bullied at school, or help them with a college application in your place? Drive you around in an autonomous car, or read your children their bedtime stories? The response from the audience was quite varying.
Her point was that each of these is already possible. Her question was a different one: “what should AI do, not what can AI do.” AI, she said, is now woven into every fabric of our lives, and she has made it her mission to humanise technology before it dehumanises us.
A choice, not an inevitability
El Kaliouby hosts a weekly podcast with AI pioneers. She recounted asking an OpenAI product leader about safety, at the same time as headlines reported AI tools going rogue. Her takeaway was that the way these products are built is a leadership and engineering choice.
She was most concerned about what she called the dangerous territory of artificial intimacy. Teenagers are turning to chatbots for support, and some young people spend five to six hours a day with these tools for companionship. She asked why the designers built no timeouts or stopgaps, and answered herself: companies aren’t incentivised to. The race, she argued, is no longer only for attention but for intimacy. She called for a new social contract between humans and AI, built on empathy and trust, and for unintended consequences to be considered from the start.

From homesickness to Emotion AI
Her own path began with homesickness. Arriving at Cambridge for her PhD, she relied on text messages and noticed how much was missing without non-verbal cues. By her account, only 7% of human connection lies in the words we use and 93% is non-verbal. That led her to build deep neural networks that recognise human emotional states at her company, Affectiva. Early systems detected something rudimentary, such as a smile or a raised eyebrow. Today, she said, they read more than 50 expressions.
The original impetus was autism: a device to help autistic children interpret non-verbal cues. Then came the question of what was financially sustainable. She described building technology to show companies, many of them in the Fortune 500, how audiences actually respond to their content. Then came a car: could this work in one? She shared two examples. One was a man who fell asleep at the wheel with his daughter in the back seat. The other was a woman driving with both hands off the wheel and two phones in use. Her team widened the camera’s view and retrained it to work with masks. It can also detect a child left in a car. She said around 100 infants die of heatstroke in US cars each year, and that the technology is now in 1.5 million vehicles.
Capital with values
El Kaliouby sold her company, Affectiva, five years ago and became an investor at Blue Tulip Ventures. Even in the early days, she said, governments wanted to use the technology for surveillance. Her company turned them down, and would not entertain any use case where people could not explicitly consent. Her advice to founders in the room: not every dollar is the same, and when times are tough you need the right investors.
The next frontier
She sees several areas of promise: AI infrastructure and wider access to AI, wearables paired with multimodal data for personalised healthcare (she has backed a group of women’s health companies), AI-driven scientific discovery, and agentic workflows that open entrepreneurship to more people. For heavy industry, such as ship welding, she sees robots taking on the most dangerous work.
Her sharper question was about the AI coworker. Most of us work in “single player mode,” each with our own tools and no shared context. Soon, she suggested, AI will sit within the org chart, in teams of humans and agents. She asked what trust, accountability and culture look like when some of the players are not human, and what compassion or integrity means for them.
What to delegate
She closed with a framework. Delegate to AI what has clear rules, high volume and easy verification. In the middle, humans decide and AI does. At the other end, humans do everything. She added an anecdote: one AI pioneer’s two children could not disagree more, with a daughter who calls AI cognitive offloading and a son who does everything with it.
Her last questions were about the moral fabric of society. People once asked their mothers for recipes. What do we lose when we ask ChatGPT instead? As the pace of change quickens, she said, it is not only the economy but the meaning of being human that is likely to mutate.