Future File

Computers that Diagnose

INTERNIST-1 is a member of a new family of “expert programs” being developed by artificial intelligence researchers. It is capable of making multiple and complex diagnoses in internal medicine. Dr. Randolph Miller and colleagues at the University of Pittsburg have developed and evaluated INTERNIST-1 by comparing it against clinicians, and case discussants. Its performance was qualitively similar to that of the clinicians, 17 correct definitive diagnoses out of 43, versus 23 for the clinicians, but definitely inferior to the discussants (29 correct). The main deficiencies in the present program are the inability to reason anatomically or temporally. The program cannot recognize problems at different stages of development or develop explanations of causality.

The impact that AI developments such as INTERNIST-1 have had on people have ranged from “that might be a handy tool someday” to visualized scenarios wherein computers take over the world. There is little doubt that as machines have been developed to handle more aid more tasks, the costs of accomplishing those tasks have fallen and the number of people directly working on those tasks has dropped, the question that libertarian futurists might consider is whether this process can continue until capital accumulation (automation and AI) reaches the point where human labor is no longer a significant part of the resource mix and the price of goods approaches zero. Singularities such as this seldom develop in stable systems—but they do develop. If this one does, how will it effect the market economy? In fact, will there even need to be an exchange of goods or will one have the capability to be a “market unto oneself”? If this singularity does not exist, what is the limiting factor?

—from “INTERNIST-1, An Experimental Computer-Based Diagnostic Consultant For General Internal Medicine,” New England Journal of Medicine, August 19, 1982

A New Electron Microscope

A scientist at the university of Chicago has designed an electron microscope capable of detecting objects one-third the size of those now observable. Dr. Albert Crewe’s design will, for the first time, make it possible to see the atomic structure of most solids. A distortion-correcting device which Dr. Crewe proposed in 1981 is a key component of the microscope, which will take three years to construct. “We are 90 percent confident that the corrector will work, ” said Dr. Crewe.

—from The NY Times News Service, January 24, 1983