23 June 2018 (Paris, France) – While I was pounding the beach in Cannes, France at the Cannes Lions International Festival of Creativity 2018, my Chief Technology Officer, Eric De Grasse was at the IEEE Conference on Computer Vision and Pattern Recognition in Salt Lake City, Utah. The event is as advertised: the premier annual computer vision event with main conference presentations and numerous co-located workshops and short courses. For exposure to state-of-the-art work in the field of computer vision and the opportunity for hands-on artificial intelligence training, there are none better.
There were over 400 papers presented and Eric has plowed through about 100. One that caught his eye was from researchers at Germany’s University of Bonn who showed the progress of research on predictive-analysis. The goal is rather startling: to predict a sequence of activities five minutes into the future. The main paper was in German (click here) but the presenters summarized their two approaches as follows:
The team tested two approaches using different types of artificial neural network: one that anticipated future actions and reflected before anticipating again, and another that built a matrix in one hit before crunching the probabilities. As you’d expect, the deeper they looked into the future, the more mistakes they made. Accuracy was over 40 percent for short forecasting periods, but then declined the more the algorithm needed to look into the future. The reflective approach did a little better than the matrix method when looking at the next 20 seconds, but the two different neural networks were equally matched when looking beyond 40 seconds. At the extreme end, the scientists discovered their trained program could correctly predict an action and its duration 3 minutes in the future roughly 15 percent of the time. That might not sound impressive, but it does establish solid ground for future artificial intelligence that could potentially develop super-human foresight.
Now, before you scream “Hold on!” yes, we’ve all read cautionary tales about the limits of this technology. For just one example, click here. I rather think the key is that this will generate more interest in predictive software and (hopefully) researchers will focus on the many positive possibilities.
And it helps focus on the nuances and difficulties of autonomous and intelligent systems. It dovetails nicely with a recent program I attended at the MIT Media Lab in Cambridge, Massachusetts … which has a long-standing history with the discipline of machine learning and AI … where we discussed the attempts to empower people with tools so they can live with artificial and extended intelligence, instead of feeling like they’re going to be replaced or destroyed by machines.
And this research also forces us to recognize that we can’t continue to measure success in purely economic terms, or to look for one-size-fits-all solutions — we have to remember that we are part of a web of complex, self-adaptive systems, which also includes the tools we use and the environments in which we live.
Eric hopes to have a more detailed analysis over the summer.