The Older Artificial Intelligence
We have been building artificial intelligences for centuries. We just stopped calling them that.
Five thousand three hundred is not a few bad actors. It is a distributed organizational output.
The legal account is correct as far as it goes: employees decided, managers pressured, executives maintained. But it cannot explain why five thousand three hundred employees, in six thousand branches across the country, over more than a decade, made the same decision. These employees were not coordinated. They did not know each other. They did not share a conspiracy. They shared a decision environment, and the decision environment produced, in each of them independently, the same behavior. The behavior was the output of the environment, not of the individuals.
From Chapter 03 · The Distributed Output
The alignment problem is not new. We have just been ignoring the first instance of it.
Banks. Agencies. Ministries. Firms. Distributed cognitive systems running on human substrate — reasoning, optimizing, producing outputs no single person inside them ever decided. We have been calling this "organization". It is the older artificial intelligence.
The unit of analysis is the room.
Not the decision. Not the decider. The decision environment.
You have been training models on humans for fifty years.
Quotas, KPIs, performance reviews; reward shaping at organizational mind
The field manual already exists.
It is called organizational theory; Luhmann, Beer, Van Doorn. We have not been reading it.
Get in touch.
Press · publishers · speaking · translation rights · advisory. Replies usually within two working days.
- Office
- Bahnhofstrasse 20 B
8272 Ermatingen, Switzerland - Direct
-
[email protected]
+41 78 227 4330 - Web
- www.nisone.ch