One of the domains that is studied at the MIT Media Lab is the use of computers in early childhood learning. Both Marvin Minsky and Seymour Papert said that the computer would only make a meaningful contribution to learning if we could find new ways to learn through the computer. Clayton Christensen from HBS shared the same view in his book on disrupting education. For example, flash cards on a computer was no better than the old paper flash cards. Khan Academy and all the MOOC programs did little to change how people learned (although they had a profound impact on the availability of good teaching).
Today on the AVC blog there is a video interview of film maker Werner Herzog. In the interview they ask Herzog if he is considering to develop movies in VR. Herzog answers that he will not do VR movies.. He says we have to look for new ways to use VR (for it to be valuable).
This concept that we should use computer technologies in new ways got me thinking about AI. Using AI to do data analysis is not all that big an improvement over what humans have done since Descartes. However, using machine learning to do pattern recognition to diagnose cancers is not something humans can do at scale. So on the one hand AI is focused on 20th century objectives with limited real benefit and on the other hand is being used for state of the art medical diagnostics of potentially enormous benefit.
The takeaway is that with new technologies we should look for new ways to use them and not fall into the trap of re-using them in old ways.