This past early March, my fellow McMaster Student Associate, Matt Brown and I had the opportunity to attend to the 2018 Esri Developer’s Summit located in Palm Springs, California. As Matt mentioned in his recent blog, this is our second time attending, all huge thanks to the ECCE Student Associates Program. Last year, I mentioned why you should go to this conference. Now, I’m going to talk about what is different this year than last year’s by emphasizing on a theme – GeoAI. Before delving into it, a brief background of what is artificial intelligence (AI).
What is AI?
AI, AI, AI. We keep hearing this term (to a point it’s becoming a buzz-term) all over the media, but what exactly is AI? Defining it is not easy as it constantly changes and has been envisioned as this constantly evolving field. Arguably, AI has been around since 1950 when Alan Turing came up the idea of “intelligent” machines via the Turing test. With the continuous development of machine learning algorithms (i.e. Artificial Neural Networks, Random Forest, and Support Vector Machines), computational power, and digital storage (think of smartphones, Internet, and IoT), these three are the vital components for AI to thrive. Since it is rapidly expanding, the AI in the near future will not be the same as the AI of the present. The most popular definition is a machine that takes actions and maximizes its chances of successfully achieving whatever goals it has at a rapid pace. Others say the true definition of AI is when it has its own consciousness, constantly learns, and makes its own decisions without human control. Think of it as the TV series – West World – or the sequels of The Matrix. Nobody knows when, and if ever, we’ll reach that phase. Certainly, it is frightening; however, if we construct AIs carefully with strict ethics involved, then this will optimally change the entire ecosystem.
Two speakers at the DevSummit spoke about GeoAI: 1) Joseph Sirosh (VP of Microsoft Cloud AI), who was the keynote; and 2) Omar Maher (Esri Advanced Analytics Practice & Data Science), an EDC presenter. Both speakers demonstrated how and what can be done when molding AI and the geographic / GIS dimension together; thus, creating the GeoAI. Some of us have already tapped into GeoAI with our research or course work projects. For instance, using the Geostatistical Analyst for kriging interpolation or applying OLS for geographic regression analyses. These are machine learning tools. The difference is to develop online platforms that can harness and process massive geospatial datasets to support near or real-time feedback for quick and effective decision-making processes. The quick and effective decision-making processes can apply to any field with massive spatial datasets. Some of the most applicable fields are logistics, farming, surveying, health, and transportation. Imagine you are a transportation planner and want to predict within 15 minutes where in the GTA will traffic congestion occur, why it’ll happen, and what can be done to prevent it. By the time it took for you to read up until this point, the GeoAI platform would have already collected millions of data points from a combination of GPS, Bluetooth, smartphone devices as well as from highway sensors, and processed it through a series of algorithms. The result is on a web GIS application and informed autonomous cars to reroute without losing ETA time. This would be the idea of how quick decision-making can be done with little compromising. Of course, this assumes a company or government agency has all the “bells and whistles” of data storage, refinement, processing, and updating. In other words, in this scenario of Intelligent Transportation Systems, it is easy said than to be done. However, we are not that far away from achieving this as most of the infrastructure has already been laid out. We just need to have more integrated systems and faster and smarter algorithms.
In health, we can create a GeoAI platform that could distribute real-time health care resources to underserved/vulnerable populations. This would potentially remove red tape processes; thus, one step closer to disease prevention, which could save nations billions/trillions of dollars annually. Overall, GeoAI can gather, manage, analyze, and predict from geographic / location-based data, and provide powerful visual insights. The past was developing AI theory. Currently and the future is about wielding this powerful tool to make significant positive changes. Hopefully, most of us in the GIS field (and Remote Sensing by using Deep Learning) will have the opportunity to get our hands dirty in GeoAI.
Platforms for GeoAI
There are multiple platforms tailored towards GeoAI. Microsoft Azure and Esri are now offering the GeoAI Data Science Virtual Machine (DSVM). This platform has the combination of cloud infrastructure, geospatial analytics and visualization to develop more intelligent applications. Additionally, GeoAI DSVM enables the client to use their GPU processes, especially when applying deep learning on imagery, for example, using TensorFlow or PyTorch. If you’re a developer and/or a data scientist, then I would recommend taking advantage of the R bridge in ArcGIS Pro, as there are plentiful machine learning (subcomponent of AI) packages that can be harnessed.
For more readings on GeoAI and AI, please take a look at the following articles and applications below: