AI is neither artificial nor intelligent. At least, not quite.
This is a provocation, not a scientific definition. Machines are manufactured, and AI systems can demonstrate capabilities we reasonably describe as intelligent. The title asks what those labels leave out.
The machinery is artificial. Much of the material is human.
Modern AI draws on human language, knowledge, writing, decisions, examples, feedback, software and culture. Behind it sit physical processors, datacentres, networks and electricity.
Calling it artificial is reasonable. But it can obscure the human contribution that makes many of its capabilities useful. One helpful way to think about it is as a means of bringing accumulated knowledge and computation to the place where work happens.
Capability is not accountability
An AI system can write, classify, summarise, analyse, work through some problems and operate tools. Those capabilities are real. They do not, by themselves, establish human-like judgement, intent, understanding or concern for consequences.
There is room for disagreement about what counts as intelligence. There should be much less ambiguity about who is responsible for a customer decision or a payment. A system’s fluent answer is not evidence that it should be trusted with either.
Calling software intelligent can encourage us to trust where we should test, or delegate where we should supervise. The practical response is to check its behaviour in the actual work, set boundaries and keep people accountable.
H2I prefers a more useful question
Rather than settling whether the software deserves the label “intelligent”, we ask: is it helpful here?
Does it reduce effort or improve a decision? Is it reliable enough for this task? Can we control what information it sees and what it does? What happens when it fails? Can the business afford to use and support it?
Sometimes the answer is a model. Sometimes it is an automation, a configuration change or a simpler process. The label matters less than the result.
We don’t need the pipe to be intelligent. We need clean water to come out of the tap.
The analogy has limits: software outputs need checks suited to their use, and no model is universally safe or accurate. That is why H2I stands for Honest + Helpful Intelligence. Honest about the limits. Helpful in the work.
AI should be like water: safe, affordable and available.
