Manifesto
The shape AI takes is a choice. Yours too. No system is neutral: every system carries its makers’ choices in it. This manifesto is the bar I hold my own work to, with every piece of advice and every system I build.
The systems I build never pass themselves off as human. No invented colleague with a smiling avatar, no "I understand how you feel". People deserve to know what they’re talking to, and trust that rests on an illusion isn’t trust.
AI may prepare, sort, suggest and speed things up. Decisions that affect people, about a job application, a complaint or a client, are made by a person. I design systems so that this human step can’t be automated away.
Choosing technology means choosing the incentives behind it. Does a tool make its money from your attention, from your data, or from the work it does for you? I choose and recommend tools whose incentives hold up: privacy-friendly, explainable, and, where possible, European and light on energy.
You can only steer what you understand. So I work without jargon, I explain what a system does and what it doesn’t, and you’ll also hear from me when AI isn’t the answer.
I don’t believe in starting big and impressive. I believe in a small system that holds up: one process, clear ground rules, and only then the next step. Technology should shape itself to the organisation, rather than the other way round.
This manifesto is written in my own words, inspired by the thinking of the Center for Humane Technology and by my own research into bias in AI. What this looks like in practice is set out in how I work with AI.
Much of what’s written above came out of what I read. I recommend these books to anyone who wants to think harder about AI and humanity.