CONCEPT
AI Deployability
The gap between what AI systems can demonstrate in controlled settings and what they can sustain in regulated, adversarial, consequence-bearing environments—the condition Miki Edelman identifies as the binding frontier of the present moment, where the capability race and the deployment reality have separated further than the technical literature suggests.
Demonstration and proof are not the same thing, and the gap between them is wider than the current AI moment acknowledges. A demonstration optimizes for surprise: the impressive output in a controlled setting, the benchmark achieved, the task completed faster and cheaper than before. Proof optimizes for the absence of surprise: the output that performs under regulatory scrutiny, survives adversarial conditions, and remains defensible six months after deployment when something goes wrong. The institutions where AI matters most—hospitals, courts, national grids, capital markets—are the institutions least patient with demonstrations and most demanding of proof. AI deployability, as a concept, names the full set of conditions a system must meet before it can function as infrastructure rather than as a slide deck: its outputs must be traceable to a reasoning chain, that chain must be articulable to an auditor, a human must be identifiably responsible for the decisions
Keep reading with YOU ON AI
Unlock the full book, 10,000+ field-guide entries, and a 1000+ thinker library. If you have a book code, register now — it takes a minute.