About this keynote
Firstly, thank you to the team at SPP Media for inviting me to Cyprus to give a keynote on the great technology race, and who wins it by 2040. After hearing some very big numbers from the speakers before me, data centres in space, millions of GPUs and trillions of dollars of AI spending, I wanted to bring it all back to just one number. One. You.
Everything is changing at once: technology, society, the environment and the economy, and it is all getting faster. The problem is that we build our lives, companies and countries for the future we can see, and most of us only see about 30% of what is here and what is coming. In 2016 I talked to Bloomberg in London about what we then called creative machines. In 2022 generative AI arrived, and no company survived first contact with it. Today we are scrambling again, this time with agentic AI, which we were calling personal digital assistants more than a decade ago. The future was being built long before most people noticed.
So we explored the 70% most people miss. A future shaped by humans using AI, and a future shaped by AI itself, which we think starts setting its own goals in earnest from about 2031. AI that has already created and pumped money through cryptocurrencies. AI that has been put in charge of mission critical cyber defence. And, yes, all of the bad news you read about, from autonomous weapons to misinformation.
But the world also has solutions, and that is the part the news rarely shows you. At the 311 Institute we work with national security agencies, governments and large organisations to build them, such as fully autonomous and adaptive cyber defence systems, while elsewhere AI is now being used to detect and watermark synthetic DNA in labs. In healthcare, predictive AI and wearables can now tell you that you are going to be ill before you feel it, which moves us from sick care to health care.
Then we looked at speed. When a new frontier model is released its capabilities reach everyone who is connected at the same moment, so your world can change in about 69 milliseconds. Today's leading models score an IQ of about 155 on Mensa tests, have around 4,000 times more general knowledge than any human, and learn around 500 million times faster. Between 2028 and 2032 we expect artificial general intelligence with an IQ of 800 or more. Are you building your company, or your country, for that?
And who wins the race? We think about 39 things decide national power, from AI, compute, energy and critical minerals to talent, standards, manufacturing and capital. On that basis our view of the leading powers in 2040 is the United States, China, India, Russia, the UK, Germany, France, Japan, South Korea and Brazil. Europe is declining on that index, but it is getting a lot right, including regulating AI, because no one wants an unregulated AI loose in their banking system. The bigger questions are who owns the AI, whether the largest technology companies are now more powerful than many governments, and what a country does if the intelligence layer it depends on can be switched off. One answer is compound AI: using several models from different places.
Finally, we came back to the one number that matters. Over the next 15 years we think the cost of intelligence falls about a thousandfold, that by 2030 AI produces around half of the world's new knowledge, and that by 2040 anyone making physical goods can make about five times more, while digital output, software, content and apps, rises about 770 times. For Cyprus, an economy of about $50 billion, the question is what happens to output, earnings and GDP when every adult can command a team of AI agents.
So my challenge to the audience was simple. This afternoon, use an AI tool to build a small piece of software, for your life, your family or your sports club. Then ask the most dangerous question there is: what else can I build? And then ask how you could sell it.
