Tracking AI's evolution

Every generation of professionals has had to absorb a new technology. Ours has artificial intelligence, and its defining feature is speed.

Medicine has been here before. The printing press spread knowledge once locked in a few libraries. The stethoscope, the X-ray and the scanner each changed what a clinician could see. The computer and the internet changed how we record, share and reach one another. None of these arrived fully formed. Each was doubted, then trialled, then absorbed across a generation, until it became part of the furniture.

A long line of disruptions

The pattern holds. A new tool appears and unsettles people. Some dismiss it; some over-promise for it. Over time the profession works out where it helps, builds the rules and habits around it, and stops noticing it is there. What first felt like an intrusion becomes how the work is done.

AI is older than it looks

Artificial intelligence is not a recent invention. Alan Turing posed his test for machine intelligence in 1950. The field was named at the Dartmouth workshop in 1956. ELIZA, the first chatbot, held conversations in the mid-1960s. A computer beat the world chess champion in 1997. For most of those seventy years the work advanced quietly, in laboratories, well out of public view.

What is different now

Recently that changed. AI arrived in everyone's hands at once and began improving month by month rather than decade by decade. A capability that looked like a research curiosity last year is in daily use this year. That compression is the real story: it leaves less time to watch, learn and adjust before the next change lands.

For anyone leading an organisation, that poses a hard question. You cannot wait for the dust to settle, because it does not settle. You cannot adopt everything either. The task is to tell the lasting changes from the passing ones.

Sorting signal from noise

Most writing about AI is built to provoke rather than inform. Real progress and marketing sit side by side and are easy to confuse. I put three questions to anything new. Does it solve a problem I have? Can I see how it reached its answer? Would I trust it with work that matters? Most things do not survive those questions. The few that do have earned a closer look.

Shifts that once took a generation now take two years.

Judgement, not prediction

I am not an engineer, and neither are most of the people I work with. They are clinicians, academics, executives and board members who must make sound decisions about AI without building it. They do not need the mathematics. They need a clear account of what a tool can and cannot do, where it tends to fail, and what it means for their people. Supplying that account is most of the work, and it is where history makes a steadier guide than hype.

What I watch for

I pay closest attention to the developments that change how decisions are made: tools that reason across large bodies of evidence, systems that can show their working, and the slow business of governing all of it. Spectacle fades. The quieter shifts, in how we think, decide and account for ourselves, are the ones that endure.

I write up what clears that bar here, in plain language, for people who would rather use AI well than talk about it. Get in touch, or read the latest updates.