Algorithms for elephants, or, halogen vs LED
Thanks again to The Batch, this positive thought-provoking news about AI in wildlife conservation.
I’d been weighing for some time the relative benefits of conservation physically (boots on the ground to stop poachers, set patrols, build out alarm systems) vs technologically (somehow implementing AI approaches that learn as the environment changes). This seems a clear example of the latter prevailing. Elephants are fitted with what at first sounds like it’d be intrusive: a device worn on/near the head, hosting a model (winners of a competition organized by Hackster.io and Smart Parks; built using dev tools from Edge Impulse; running on hardware from Institute Irnas and Avnet) that
- either recognises human sounds picked up by the collar’s microphone and cross-references them with GPS coordinates to detect possible poachers.
- or uses data from the collar’s accelerometer to determine when elephants are eating, sleeping, or running.
- or analyzes motions and vibrations of an elephant’s trunk for signs of distress from human interaction and alerts people nearby.
- or performs more than one task in combination: interpret elephant activity by those means above, as well as alert rangers to sounds of weapons commonly used by poachers (the AK-47 has a signature sound).
Other ML applications benefitting conservation efforts and, perhaps less intrusively (perhaps there’s evidence available somewhere as to dis/comfort experienced by elephants wearing those collars described above, but I haven’t been able to find to find it yet), we can run the following back in the lab:
- a model called PAWS suggests optimal patrol routes to help park rangers in Cambodia intercept poachers.
- a model from Google Earth recognizes 614 species and classifies 3.6 million images an hour to aid quantifying of engangering activities.
I’m not advocating for robot bees any time soon but if we can implement AI to help tackle the tangible problems in conservation (like poaching) then we must be doing something right. It’s like when halogen bulbs started to be outlawed in the UK (sparking a run on any stores stocking them) only for us to realise the technologically more sensible, informed approach, would have been, and is to this day, to switch more quickly from fossils to renewables. Indeed perhaps it’s the case that sticking with halogen all along would’ve meant increased demand and motivation for getting the wider country & industry to make that switch. It all became academic shortly after the switch however, due to the covid pandemic and all trimmings in terms of social upheaval as folks swapped the office and the corner coffeeshop for a remote working life spent at home with presumably fewer lightbulbs firing across the country on any given day (how many LEDs does it take to power the lobby & lifts of each office block in The Square Mile?)…