A bubble inside a longer cycle

Nicolas de Santis has seen technology cycles before. An early dot-com founder who co-founded Opodo and helped launch one of the first digital currencies, he points out that AI has been part of computer science courses for around 30 years. Asked directly whether the current boom is a bubble, he answers yes, explaining that bubbles form in many industries because everyone wants to make something out of a new opportunity. Howard Bilton, who grew the Sovereign Group to 600 people, sees AI as a source of efficiency that still needs practical expertise.

Keyu Jin offers a structural view of where technology comes from. In her account of China's innovation flywheel, innovation is embedded in production and improves through scale and learning by doing. Electric vehicles, she notes, grew out of consumer electronics through battery expertise, and Beijing sets strategic direction in areas such as AI and renewables while local governments implement it. The state may seed a sector, but Jin argues that private competition does the selecting, with hundreds of companies entering a market and being reduced to a handful. Marco Poon links AI to physical infrastructure, seeing a global need to overhaul power grids to meet its energy demands.

The tutor that does the work of thousands

Shield Xiao Dun has lived through the disruption he now builds. His first company, 17 EdTech, grew to 20,000 staff by using mobile phones and faster networks to reach learners far from China's first-tier cities. His new company, in Palo Alto, builds AI tutors that guide each learner in a personalised way. Because generative AI began with language models, he explains, a tutor can answer even a bizarre question patiently, at any hour, and follow a student's curiosity as far as it goes. Michael Macfarlane notes that AI now does what 20,000 people once did.

Xiao Dun expects assessment to change as well. Interviews, in his experience, are unreliable, and judging someone's output over time is far better, though it was too labour-intensive for schools before AI. Now that AI can judge essays and understand video reasonably well, he expects change to come from outside existing institutions.

So I think there will be a new mastery based system coming out of this new generation or new era.

— Shield Xiao Dun, Wealth Office

He takes the risks seriously. His worst case is a population kept consuming without ever growing. His second is a small group owning the compute and the best models and making every strategic decision. He prefers a society shaped like a cactus, with many branches and many peaks, where hard work is rewarded and value is measured in more than money. He admires technology that the wealthiest and the poorest use alike, and hopes his tutors will reach learners on that basis.

Trust under attack

For family offices, the most immediate effect of AI may be on security. David Ferbrache OBE, Managing Director of Beyond Blue, says AI helps criminals scan many systems quickly and research targets thoroughly, producing credible phishing lures and, at times, simulated video and audio. Impersonation, in his account, is increasingly the source of damage. He describes law and accountancy firms being compromised so that attackers can email clients with the firms' credibility. He also says the time between initial compromise and attack is shortening. When Macfarlane calls this a hijacking of trust, Ferbrache agrees.

It's a hijacking of trust, exactly.

— David Ferbrache OBE, Wealth Office

Lord Bernard Hogan-Howe, former Commissioner of the Metropolitan Police in London, adds that AI is starting to defeat biometric safeguards. He notes that software can identify where an image was taken, which links the next generation's online oversharing to physical risks such as kidnap. Neither guest argues for withdrawing from social media. Both describe good security as unobtrusive, allowing a family to travel, associate and do business while the risk is managed around them.

Craft, plumbing and the meaning of value

In the art world, guests expect technology to change what is scarce. Pearl Lam argues that AI could make conceptual art unnecessary, since a prompt can produce a concept and the look of an installation. Painting and sculpture, long looked down on by art schools that teach concept, may in her view be valued very differently in the next 10 to 20 years.

One thing, I always say that if you have an AI, you don't need conceptual art because you just type what you want and then everything will come out.

— Pearl Lam, Wealth Office

Lam sees homogeneity as a wider problem, driven less by AI than by clients who commission what feels safe, and argues for celebrating difference. Bigna Pfenninger, who says she tokenised a Picasso in Switzerland in 2021, treats tokenisation as infrastructure. Fractional ownership is old, she notes, as five brothers who inherit a work already know, and tokenisation is a cheaper, more effective way to divide it. She expects it to become unremarkable and sees clearer value for museums, donorship and joint ownership, while doubting that stock-exchange listings will bring depth as well as trading volume. Lam is similarly sceptical about shares in art, recalling an attempt in China around 2007 that she says failed, and asking who would manage such a work and which law would govern it.

Lady Linda Wong Davies finds AI enthralling but believes it will lack humanity. Across these conversations, technology appears as a set of pressures, each changing where value and risk sit. Guests see it widening access in education, sharpening threats in security and shifting worth in art towards what machines find hardest to supply.