AI Is Not the Great Equaliser. It Is the Great Divider.

This past week, I was invited by the South China Morning Post to speak about Hong Kong's digitalisation. But I could not help sharing my views on our education system as well. One of the greatest myths I have heard since the birth of ChatGPT is that AI is the great equaliser. It is not.

The invention of the wheel did not make everyone Michael Schumacher, just as the invention of football did not make everyone Cristiano Ronaldo. AI will amplify the capabilities of those who use it effectively, while those who do not will be left eating their dust. Given that we are, by nature, creatures of inertia, a significant portion of the current generation will be left behind. That is simply the reality.

According to Microsoft's AI Economy Institute, more than 1.2 billion people have used AI tools in less than three years, yet adoption in the Global North is roughly twice that of the Global South. But the problem is even more nuanced than that.

Even within my own team, I would say that only half of us are genuinely effective at using AI to increase productivity. The other half, despite having access to exactly the same tools, have failed to achieve similar improvements. I count myself among this second half and constantly feel as though I am being left behind. If this is happening within an AI-focused venture capital firm, the situation elsewhere is likely to be far more dismal.

The real question, therefore, is not whether AI is fair. It is whether we are willing to become AI-native. People who do not use AI will not simply move a little more slowly. They will be priced out of opportunities and excluded from the new language of work.

Here is my grievance as a mother: our education system is still optimised for a different century.

It was built like a pyramid: standardise the many, filter out the few, and then crown a small class of experts who become bankers, lawyers, and doctors. That model produced reliable professionals for an assembly-line economy. But we are no longer in the 19th century. In an era when intelligence is becoming abundant, the scarce skill is the ability to connect the dots across different disciplines.

What we desperately need now is a more Montessori-style system: one that trains curiosity, rewards exploration, embraces failure, and encourages constant iteration.

My parents did not know what Montessori was, but they were kind enough to leave me alone. I moved from studying hotel management to physics, and from working in technical support to Wall Street, followed by an AI start-up and, now, venture capital. Somehow, it all worked out ok, not because reducing chicken stock prepared me to invest in silicon chips, but because learning across different domains creates its own asymmetric advantage over time.

I am therefore highly sceptical of any education model that mistakes compliance for intelligence.

That same mindset is reflected in how we think about AI infrastructure. When most people say "invest in AI infrastructure," they mean more land, more power, and more chips. But the real bottlenecks are energy, cooling, water, land, and grid capacity: constraints associated with a Kardashev Type I civilisation. Once you recognise those constraints, you begin to ask whether Earth is necessarily the right place for all future compute. Elon Musk has helped validate the broader orbital-computing sector, but Starcloud, one of our portfolio companies, has already become the first company to train a large language model in space, an early step towards Type II energy infrastructure.

Bringing the discussion back home, this is the mindset governments should embrace when thinking about digitalisation.

Digitalisation is not simply about moving forms online or deploying another enterprise dashboard. It is about building a society capable of using AI to think better and invent more boldly.

Interestingly, the countries leading AI adoption are not necessarily the largest builders of frontier models. In addition to China and the United States, Microsoft identifies the UAE, Singapore, Norway, and Ireland as AI-adoption leaders. Their experience suggests that access, education, and, above all, policy coordination matter alongside technical capability.

We cannot have 20th-century policies governing 21st-century technology.

At 3C, the "C" stands for curiosity. To me, curiosity is not a soft value; it is hard infrastructure for the mind. In the AI era, curiosity compounds faster than credentials.

AI is not the great equaliser. It is the great divider.

Good luck to all the parents out there.