I saw a very interesting TEDx chat by Jeremy Howard where he talked about the current state of machine learning and what computers are capable of today. Howard is the founder and CEO of Enlitic a medical company that “uses recent advances in machine learning to make medical diagnostics faster, more accurate, and more accessible.” His talk was called “The wonderful and terrifying implications of computers that can learn” and was delivered last month at TEDx Brussels.
As a cataloguer and metadata specialist what really caught my attention here was Howard’s classification example where he models the medical diagnostic process by sorting through 1.5 million images of cars. The “extraordinary algorithm called deep learning”* was used to accomplish this task in 15 minutes something that might take “a team of five or six people about seven years.”
It’s wonderful to watch. And I marvel at this process and wonder if this technique might be used as a way to navigate through large linked data sets. The deep learning algorithm operates in a “16,000-dimensional space” which does a lot of the heavy pattern recognition lifting. The system then learns what it is we are looking for through hints and nudges that we provide based on the resulting collections of images.
Howard sees this technique as one way to solve the “lack of medical expertise in the world.” The World Economic Forum indicates that it would take about “300 years to train enough people” so that there are enough medical personnel in the developing world.
Working with the deep learning system diagnostic process can be sped up leaving the analysis of the medical data to the computer.
He’s worried though because computers are now capable of reading, writing, seeing, speaking and listening, and integrating knowledge. Which means computers can now drive cars, prepare food, diagnose disease, find legal precedents, etc. What does that mean to a world where a high percentage of the services humans now provide are “also the exact things that computers have just learned how to do”?
He provides a comparison of the social disruption caused during the Industrial Revolution to this new “Machine Learning Revolution”:
“We have seen this once before, of course. In the Industrial Revolution, we saw a step change in capability thanks to engines. The thing is, though, that after a while, things flattened out. There was social disruption, but once engines were used to generate power in all the situations, things really settled down. The Machine Learning Revolution is going to be very different from the Industrial Revolution, because the Machine Learning Revolution, it never settles down. The better computers get at intellectual activities, the more they can build better computers to be better at intellectual capabilities, so this is going to be a kind of change that the world has actually never experienced before, so your previous understanding of what’s possible is different.”
A couple of weeks ago Howard participated in a discussion on Reddit. In it he mentions the image sorting demo he used above and how it may lead to an evolving relationship between humans and computers:
“The demo that I showed at TEDx gives an example of how to create a classification algorithm from scratch without requiring any code. Over time, we will continue to build higher and higher levels of abstraction, so that the computer does more and more of the work. That way, the human is providing higher and higher level instructions as to what they want the computer to do, rather than telling the computer how to do it.”
He also talks about how this may all lead to “technological unemployment” but the jobs that are left will be more “intellectually intensive.” This type of transformation can also be seen in the world of library cataloguing where the description of a resource can be passed along the information supply chain leaving only the subject analysis of the intellectual content to the cataloguer. Something talked about for years but not yet fully realized.
Howard concludes the Reddit discussion with a comment very reminiscent of the Robert A. Heinlein story, “For Us, The Living: A Comedy of Customs” where “individuals are able to choose whether or not to accept a job” and there are “short working hours and high wages paid to employees”:
“If we can get to the point where we are, on the whole, past scarcity, then they may not look like ‘jobs’ that we have today; it would be closer to the kinds of things that we do in our leisure time. This would include creating and consuming art, participating in and watching sports, having philosophical discussions, getting together with friends, and so forth.”
He concludes the TEDx talk with this call to begin considering the potential social changes this revolution will bring:
“Computers right now can do the things that humans spend most of their time being paid to do, so now’s the time to start thinking about how we’re going to adjust our social structures and economic structures to be aware of this new reality.”
This is a fantastic summation of the current machine learning environment that is developing. Take twenty minutes and get a glimpse of the “new reality” that’s already here.
* an algorithm modelled on how the human brain works developed by Geoffery Hinton, at the University of Toronto, for more see this New York Times article