eMote: AI and Change Musings

The first three pieces in this series followed the gravity. In AI Is a Black Hole With a Bubble at the Edge , the concern was capital: real technology, real usefulness, but an ecosystem increasingly full of circular deals, huge valuations and heroic assumptions about future demand. In When the Mine Builds Its Own Shovels , the concern moved to machinery: OpenAI’s Jalapeño chip, Nvidia’s moat and the moment the biggest miners decided they didn’t just want to buy shovels. They wanted to build them. Then came the infrastructure question. The mine needs a grid. AI does not live in “the cloud”. It lives in data centres, power agreements, cooling systems, planning applications, water demands, debt structures and local consequences. So where does that leave the rest of us? Not the hyperscalers. Not the chip companies. Not the model labs with cathedral-sized ambitions and balance sheets held together by confidence, capital and electricity. I mean councils. NHS trusts. univers...

Cloud is a comforting word. Soft, white, weightless. Something above us, not beneath us. Something that floats. It may be the most successful misdirection in modern technology. AI does not live in the cloud. It lives in buildings. Very large buildings. It lives in data centres with substations, cooling systems, fibre routes, planning applications, security fencing, backup generators, water requirements, debt structures and local objections. It lives where electricity arrives, heat leaves and the financial model hopes both keep behaving themselves. That is why the third turn in this AI black-hole story has to be physical. The first article asked whether AI is a bubble or a black hole. My answer was yes, inconveniently. It is a real technological force surrounded by speculative financial behaviour. The second article looked at OpenAI’s Jalapeño chip and the moment the mine starts building its own shovels. Chips matter because whoever controls compute controls the economics of...

The first stage of a gold rush is noise. Claims, rumours, men with maps, men without maps, money arriving faster than judgement and everyone pretending they had seen it coming. The second stage is machinery. At some point, the serious players stop merely digging. They start controlling the tools, the water, the transport, the crushing equipment, the finance and the town around the mine. The prospector becomes an operator. The shovel becomes strategy. That is why OpenAI’s Jalapeño chip matters. Not because it is the end of Nvidia. It probably isn’t. That would be far too neat, and history rarely organises itself for a headline. Nvidia is not just a chip company in the simple sense. It has GPUs, networking, CUDA, libraries, rack systems, supply relationships, developer trust and the muscle memory of a whole industry. It is not a man selling spades beside a river. It is the railway, foundry, shovel-maker and toll road all at once. But Jalapeño changes the question. The questio...

  AI, agency and the next experiment in democracy For most of human history, words couldn't actually do anything. They could make us do things. A king could issue an order, a priest could pronounce a marriage, Parliament could pass a law and a bank could refuse you a mortgage. But somewhere between the words and the consequence there was normally a human being. Someone had to hear them, understand them, decide what they meant and act. That's beginning to change. I was listening recently to Yuval Noah Harari, the author of Sapiens , talking about artificial intelligence. He makes a distinction between a tool and an agent. A knife is a tool. It doesn't decide whether to cut a tomato or stab somebody. An agent can decide what to do. AI, he argues, is becoming the latter. I'm not convinced the distinction is quite that clean. I've been circling around this problem for a while, particularly in my writing about AI agents. In RACI for Machines I argued that the interesti...

I think I’ve finally worked out why people can look at the same AI and come away with completely different conclusions. Some see the beginnings of another intelligence. Some see a glorified autocomplete. Some see an existential threat. Some see an office tool. I suspect the truth is both less dramatic and more interesting. Perhaps AI’s great trick isn’t that it thinks. Perhaps it simply helps us avoid ignorance at speed . That sounds rather underwhelming compared with artificial general intelligence, machine consciousness and the imminent replacement of humanity. But the more I use these systems, the more I think it may be the more important idea. Ask an AI about something you know extremely well and the magic often disappears. You read the answer and think: yes, fair enough. Mostly right. A bit obvious. Occasionally wrong. Sometimes very confidently wrong. But then ask it about something just outside your own field. Suddenly it becomes much more interesting. You might star...

What exactly is a political party for? That sounds like the opening to a Count Binface manifesto. Perhaps it should. Count Binface has managed the impressive feat of standing for serious public office while proposing policies that mix nonsense with recognisable frustration. The joke works because buried among the absurdity are things people actually care about. But there is another joke underneath it. How coherent are the manifestos of the serious parties? We pretend political beliefs naturally arrive in packages. If I believe A about taxation, I should probably believe B about immigration, C about housing, D about the environment and E about foreign policy. Why? Human beings aren't really like that. I can agree with you strongly about AI regulation and disagree completely about taxation. We may agree again about housing. Then discover entirely different views on transport. Political parties take this mess and bundle it. Artificial Personal Intelligence c...

There is an obvious problem with giving artificial intelligence political power. You can make as many AIs as you like. Democracy, by contrast, has spent centuries struggling towards the principle that political authority belongs to people, not property, wealth, corporations or whoever owns the biggest megaphone. Then along comes agentic AI. I get one agent. A company gets ten thousand. A political campaign gets a million. Someone with enough compute gets an army. So perhaps the first rule of Artificial Personal Intelligence should be surprisingly old-fashioned: One person. One democratic identity. One vote. The AI doesn't get a vote. You do. It represents you. That distinction may turn out to be constitutional. The £10 problem Imagine a regulation costs one company £100 million. That company notices. It hires lawyers, economists, lobbyists and public affairs specialists. It commissions research and follows every amendment through Parliament. Now imagi...