Pattern, at scale

[AI-assisted for speed.]

AI makes a lot of sense to me because it works, at least conceptually, the way I do. It identifies patterns from accumulated relationships and experiences, then applies them in entirely different circumstances (without necessarily caring what any of it is about). The pattern becomes something in its own right, independent of the substance that produced it.

I’ve written before about my preoccupation with form. Design, architecture, rhetoric, etiquette, persuasion. These aren’t separate interests. They’re expressions of the same underlying principle: form doesn’t follow function. Form dictates, defines, and enables function.

One of my favourite tells is how people describe themselves as being in sales, rather than being in whatever they happen to sell. The product matters, obviously, but the mechanics of persuasion are remarkably consistent. Many fundraisers approach their work as academics, advocates, or lawyers, convinced that the merits of their argument ought to be sufficient to persuade someone. I’ve spent much of my career trying to get folks to understand that persuasion is a discipline in its own right, not simply the presentation of a compelling case.

Which brings me back to AI.

There were, broadly speaking, two competing approaches to artificial intelligence. One attempted to construct intelligence by explicitly defining the world. A stop sign is red, has eight sides, contains the word STOP, and so forth. Give a machine enough definitions and rules, and it should be able to reason about things.

The other approach was to build systems that learned to recognize patterns through repeated exposure to enormous amounts of data. Rather than being taught what things meant, the neural network learned their shapes, structures, and relationships. It could then recognize those patterns in circumstances it had never encountered before, and produce remarkably accurate results without necessarily understanding the substance of what it was dealing with. It’s this approach that produced the extraordinary advances of recent years.

Yet one more data point that salespeople run the world. And the primacy of pattern+shape.

Neural networks prove that form can generate sophisticated outcomes without understanding. And maybe challenge the hierarchy of where meaning lies.

Leave a comment