Currently reading:
Dave Gershgorn (27 August 2017) 'The age of AI surveillance is here', Quartz.
James Vincent (23 January 2018) 'Artificial Intelligence is going to Supercharge Surveillance: What happens when digital eyes get the brains to match', Verge.
Vincent notes that video surveillance & CCTV technology are often described as 'digital eyes', but in truth are closer to 'portholes' in that there needs to be a driver behind the technology that is orienting and understanding the feed of images being created by the sensors. However, where these surveillance tools are now powered by artificial intelligent software, Gershgorn notes that the oft-thought prohibitive nature of trawling through volumes of surveillance footage to identify a particular individual, activity or event is becoming accessible to (still to very well-resourced) public and private sector entities.
We've reached the tipping point where AI is now being brought (at perhaps an accelerated pace) into the fold of urban crime prevention strategies. The complexity of images - 'millions of pixels that form unique patterns... [once] too complicated for hand-coded algorithms to reliably work' have been addressed by advances in deep learning. Described as 'deep neural networks', these artificial-intelligent surveillance systems can, where given a sufficiently large body of reference images, develop mathematical patterns that ably identify common characteristics between images. The reducing the rate of error is now allowing these autonomous systems to be 'trusted'.
For now, these technologies are primarily applied to facial recognition purposes, and surveillance of human behaviour. This likely arises from general public acceptance of surveillance for crime prevention purposes - however, I'm suggesting that the technology cold easily be re-purposed to the construction industry and verify (without human intervention and via machine vision) if certain conditions precedent or subsequent have been met, or to certify progress of works on site.
A number of examples of AI surveillance are cited:
- Facebook is seeking to apply the technology to understand who is in a in a video livestream and to be able to ascertain what the subject of the video is doing. Director of applied machine learning Joaquin Quinonero Candela states, 'ideally, Facebook would understand what's happening in every live video, in order to be able to curate a personalised video channel for users'.
- Baidu is employing facial recognition software as an alternative to ticketing infrastructure at large events. Such similar use of facial recognition as a form of admission progress has been tested at Charles de Gualle airport (France) and a programme piloted in Japan in 2016.
- Axon and Motorola has expressed plans to infuse artificial intelligence into its bodycam products, that, when coupled with facial recognition capabilities may aid in the search of, for example, missing persons.
