Turing Test Passed in Gaming, Flying Car Arrives, Upload Predictions Surprise, Ubiquitous Digital Pathology Cannot Be Far Behind! – Kim Solez, M.D.
It’s tough to make predictions, especially about the future.
I am a pathologist who spends much of my time writing about and teaching about the technological future. Keith Kaplan has asked me to contribute some guest blogs in this space.
The future has a way of coming at us sideways. While on the one hand, it is very hard to predict, on the other, many of the most well-known predictions end up happening. All too often, however, the future unrolls in a way, or in timeline that that is very different from what was originally envisioned.
Three examples of this are the arrival in 2012 of (1) two bots that seem to fulfill the gaming equivalent of the Turing test, (2) a production model of a flying car, and (3) economist Robin Hanson’s surprising description of a future world of mind uploads “A Tsunami of Life: The extraordinary society of emulated minds” at the just concluded Singularity Summit 2012 in San Francisco. In this blog piece, I will reflect on these three events in the context of the University of Alberta course I teach on Technology and the Future of Medicine, (introductory lecture here) and discuss some future implications for digital pathology.
As it happens 2012 is the centennial of artificial intelligence pioneer Alan Turing’s birth. It is also the year in which two shooter bots, nicknamed UT^2 and MirrorBot, appear to have passed the Turing Test in the gaming world.
What is the Turing test? In his 1950 paper Computing Machinery and Intelligence, which poses the question “Can machines think?” Turing predicted that one day computers would surpass us in intelligence. He described what we now call the Turing Test in which a human observer cannot tell the difference in typed responses to questions from a computer and a human who are both hidden from view. This is regarded as an important future milestone in computer intelligence.
By extrapolating from the exponential growth of technology over several decades, futurist Ray Kurzweil has predicted in his 2005 book The Singularity Is Near that Turing Test-capable computers would be manufactured by 2029, but now it looks like something comparable has been achieved in the somewhat more limited gaming world in 2012. The Singularity referred to by Kurzweil is that point in the future, predicted to be 2045, where machines become smarter than human beings and begin to control the agenda of the world. The presence of Turing Test-capable computers in some sense in 2012 moves us closer to the Singularity.
The winning shooter bots UT^2 and MirrorBot were able to convince a panel of judges that it was more human-like than half the humans it competed against. In this sense they did not just pass the Turing test, they actually considerably exceeded its requirements.
To some readers the gaming world may seem to be something quite distinct from normal life, but to most young people today video games are normal life. Ninety-seven percent of children play video games and they are also a daily activity of a substantial number of teens, 20-somethings and 30-somethings today. In keeping with the theme of the future coming at us sideways, this important artificial intelligence milestone in gaming is not exactly what was envisioned originally when the Turing test was described, but it is highly significant and relevant to the future of humanity nonetheless.
The question Where’s my flying car? has been a standard putdown for those who put too much faith in technologic predictions and the advance of science. Practical commercially viable flying cars were predicted to appear by the year 2000 but were nowhere in sight at that time. Then recently just at the time when many were resigned to them never appearing, the Terrafugia Transition arrived. This vehicle is limited by the fact that it can only take off and land at airports, and therefore is nowhere near as versatile as the fictional flying cars found for instance in the 1997 movie, The Fifth Element and probably very different from most conceptions of a flying car. So, once again, it will be the future coming at us sideways, when the production model rolls off the assembly line later in this year, 2012. Nevertheless, it too represents a very important advance.
The third example of the future coming at us sideways has to do with mind uploads. Most of us have very little idea how a society of human minds uploaded to computers would work. The one thing we probably all can agree on is that such a world would be very unfamiliar and unlike anything we have encountered before. It was very surprising then to find a familiar element in noted economist Robin Hanson’s presentation “A Tsunami of Life: The extraordinary society of emulated minds” at the just concluded Singularity Summit 2012 in San Francisco. In his presentation (minutes 9:10 to 10:13) Hanson predicted that fewer than 300 highly unusual humans with very special capabilities would be uploaded, but these few hundred unique minds, famous and known to everyone, would be replicated trillions of times all over the world so that every community would be like a small town with forager sociality from olden days where everyone knew everyone else. This idea that every uploaded mind you encountered in the future would be of a person you already “knew” is astounding and most unexpected, once again the future coming at you sideways.
Whatever ones preconceptions about the roll out of ubiquitous digital pathology, I would imagine that it will also come at us sideways. For instance, is it your expectation that all pathology slides will be digitized soon in the developed world?
This is a subset of the question of whether all information in the world should be digitized. On the one hand speed of digitization is increasing exponentially and storage space is becoming cheaper and cheaper. These trends were already apparent in 2005 and are even much more obvious today. On the other hand Australian AI researcher Marcus Hutter, who teaches in the University of Alberta course on Technology and the Future of Medicine, (lecture here) has pointed out that in some interpretations of information theory an enormous library containing all possible books would be just as worthless as one containing no books at all. Without some filtering, searching, and selection information has no value. It is also interesting how you make the selection of knowledge to be inputted. On the one hand much of the content in the Internet of Things, data for instance on refrigerator compressor or furnace starts would be of no value to most of us, and would just dilute the rest. On the other hand Thing Knowledge, the human knowledge contained in the design of complex machines, is some of the most valuable knowledge we possess as human beings, although there is a need for further sorting and categorization of that knowledge as well.
So where does this leave us with digital pathology? Just as in the enormous library containing all possible books it could be impossible to find Shakespeare, so also in the “library” of microscope slides it is probably impractical ultimately to digitize all of them. How we will select the cases and slides to be digitized in the future and the technology and algorithms employed will probably contain surprises as great as the surprises implicit in the Turing test, flying car, and mild upload vignettes above.
One element of my 2005 meeting with iconic Canadian poet singer/songwriter Leonard Cohen which I left out of the published report is his comment on how useful forgetting is. “Imagine how cluttered your life would become if you could remember everything!”, he said. Do we need to remember everything? Forgetting simplifies and there are probably types of “forgetting” that make sense even in a business context. Extrapolating this to the day jobs of those reading this blog, it will not make sense to digitize every slide or every case. Some cases will just require a simple placeholder image, nothing more…
For the foreseeable future the regulatory requirement will be retention of glass slides. Perhaps for the less interesting cases, all that is required is the glass slides or at most one low resolution digital image of one representative slide and nothing more. The collection of those low res image will be something familiar and low tech, like a throwback to the 60’s, analogous to the small town where everyone knows everyone else, something we may think back to nostalgically and thought we would never see again in the future.
Whatever happens in the roll out of ubiquitous digital pathology, the surprises will involve both the unfamiliar and the familiar, and often come from most unexpected directions!
Stay tuned to this space for more surprising speculations about the future.