Stephen Thaler’s Imagination machine has improved everything from a toothbrush to warheads. It has even composed a music album called “Songs of the Neurons”. How does this work? He uses an artificial neural network. These networks “dream” based on what they know. Random “noise” triggers the machines to dream. One neural network is stimulated via computational ‘heat”. While another is a “perceptron” watching the stream of ideas. The perceptron micromanages the heat in order to generate betters ideas. These “imagitron” networks are not written by humans but are self-assembled.
How much should we trust these AI machines? Should we let them diagnose illnesses? Machines can store and know more about a wide variety of illnesses. They should be better at diagnosing illness. If we let all the doctors go away and rely on the AI machines, then what happens if one of the AI machines produces invalid responses? Do we even notice?
The technology to implement this is here. But we need to deal with the moral implications before proceeding.
http://wfs.org/May-June09/Thalerpage.htm
2 comments:
A possible approach is to follow what we are doing with airliners. Modern aircraft can takeoff, navigate en-route, and land by themselves. However, we have a trained human pilot there to monitor the system for errors and to override the automated system if necessary. A similar approach could work for medicine where automated systems propose diagnoses and human doctors monitor them to screen out erroneous findings. However, this is not a perfect system. For one thing, relegating the human to being a backup for the automated system devalues the human and makes it difficult to develop experience.
With regards to the AI machines and their diagnostic limits.
It is not so much what is observed as what remains unseen that increases the value of human input in the diagnostic process.
Extrapolating beyond the evidence, questioning the data and interacting with the patient help medical professionals discover what is unknown.
Post a Comment