Every answer a chatbot has ever given you was built one word at a time. Every word, a guess. Now the world is betting hundreds of billions of dollars a year that those guesses will add up to intelligence.
In 2012, the neural network that kicked off the modern AI boom was trained on two gaming
graphics cards over roughly a long weekend. By 2026, five American companies were on course to spend more than $700 billion in a single year on the chips, buildings and power needed to push that same idea as far as it will go. Measured against the economy paying for it, that is close to the biggest peacetime industrial mobilisation in American history.
The bet behind it is simple: make the machine big enough, feed it enough of what humanity has ever written, and somewhere along the way prediction will turn into understanding. The Most Expensive Guess in History argues that the bet is wrong. And because AI is already woven into how businesses run, how governments operate and how militaries plan, the cost of being wrong won't stay where it starts. It will travel.
This is not a doom book. The author has held research posts in machine learning, has spent more than two decades working in the field, and now builds AI start-ups: an insider who believes AI will be a genuine force for good, and who can see that we are not building it on solid ground. Written from inside the machinery, this is a clear, plain-English guide to what is really happening under the bonnet, and to the warning lights most people are missing.
Inside, you'll discover: - How machines actually learned to talk, explained without jargon, and why a system that sounds intelligent isn't necessarily intelligent at all
- Why a chipmaker investing up to $100 billion in a customer that is free to spend it on the chipmaker's own chips should give every investor pause
- The hidden physical cost of AI: the water, the power, the land, and the towns suing over a hum they can hear from a mile away
- Who really holds the off switch once AI becomes a matter of national security, and why one small island sits at the centre of it all
- Whether this is a bubble, and what bubbles tend to leave behind when they burst
- What it would truly mean for a machine to understand, or even become conscious, and who would get to decide what happens next
- Why electricity, not algorithms, may turn out to be the limit that matters most
- What changes for all of us when AI stops living in a chat window and starts walking around in a body
Written for anyone who uses AI, invests in it, works alongside it or lives near where it is built (which, increasingly, is everyone), The Most Expensive Guess in History is accessible, evidence-based, gently funny and lightly illustrated. It asks the questions the AI industry would rather you didn't, and by the last page, its title will sound less like a provocation and more like a fairly accurate description.
