What do Black Moth Vision Systems, Retina Visions, Vaisala RoadAI, Frontline Data Systems, Mobileye, RoadBotics and EagleSoft have in common? Deep learning. Yet that is not the whole story.
Local Council missed out in the NSW Government’s $50 million Fixing Local Roads Pothole Repair Round. Monies went to 94 regional councils, in dire need to repair flood- ravaged roads. The Greater Sydney focus is developing artificial intelligence to machine learn about edge detection and spectral segmentation of potholes, high resolution cameras, 3D stereo image analysis, GPS locators, vibration acceleration gyroscopes, and 5G Internet of Things (IoT).
On 28 November 2022, Digital NSW, a Whole of Government initiative of NSW Customer Service, announced Asset AI® Digital Recovery of local roads to make NSW a ‘Safer Place’. The Institute of Public Works Engineering Australasia with a local council in western Sydney completed phase one trial. It now moves to final phase completion in early 2024. Our local Council may participate from late 2023.
Deepfake takes its name from swapping human faces in video and digital content to make realistic-looking fake media. Anyone with an iPhone app can make deepfakes. A portmanteau of words: Deep Learning by machine robots, Artificial Intelligence, Neural Networks, and, of course, faking it. Generative Adversarial Networks (GANs) detect and improve flaws in deepfakes with multiple neural networking. We should be very concerned by deepfakes filling the web.
Today our Council sends out traffic engineers to audit roads. They drive around assessing pothole patches for complete road resurface. Council takes four years to complete one whole audit cycle. During these four years, the roads deteriorate beyond what was audited. By the time a road maintenance crew comes along to spade cold mix asphalt emulsion into potholes, it is simply going to washout next heavy downpour. Additionally, the mix is big batched at the depot. It quickly deteriorates in the back of a Council dump-truck. A technique unchanged in the last 50 years. Naturally, stupid.
Deep Learning deploys a machine attached to a public bus or garbage truck to auto-mate the pothole audit. Conceptually, it is quicker. GANs also improve the quality of the machine learning. It does not however fix the pot-hole. Well might Premier Perrottet react last November with ‘Porque no los dos?’ (Why not both?), to NSW Roads Minister Natalie Ward’s Liberal Party preselection defeat in the safe seat of Davidson for the upcoming March State election. In the peculiar situation where western Sydney bogans eat tacos, both hard and soft shell, with their limited knowledge of Spanish, they may instead retort, “You good? Or nah?”
Yet where literally there is rain over a wet (potholed) surface, deep learning is ‘Llueve sobre mojado’ (meaning, to flog a dead horse). What local roads need now is heavy equipment and skilled labour. We are talking about pothole pro combo road cutter / sweeper bucket / hydraulic-cropping machines; spray injected cold mix modified asphalt; and volumetric thermalisation on-site of small-batched polymer-modified asphalt.
Until Council’s pothole repairs use innovative permanent physical techniques, the digital Asset AI® is deeply faking it. Artificial intelligence is no substitute for natural stupidity.