BOSTON - Should the world be worried that some of the best-known leaders and innovators of artificial intelligence (AI) are warning that it threatens human existence, even as Microsoft is claiming that its AI is developing distinctly human modes of thought? Or should people take solace from those who insist that AI is just another tool, and one that is not good at many things, or even all that good at those few things, either?

The AI-as-weapon-of-mass-destruction and AI-as-smart-hammer narratives are both predicated on the common assumption that AI operates - and will continue to operate - as an entity that is entirely distinct from humans. Whether AI is a clever tool or a rival to humans, the notion here is that AI will advance on a parallel track that is disconnected from who humans are, what they do, and what they want. As it does so, it will be increasingly inserted into people’s current ways of working, either serving humans or taking their place.

One way that that assumption of separate and parallel AI development plays out is in the ways people consider its healthcare applications. The most common approach to exploring and discussing medical AI is to turn it loose on images or other patient data to look for abnormalities or make diagnoses, and then compare its performance to that of clinicians. These discussions usually move off into speculations on how useful these tools could become as they improve, serving as a