The following is a continuation of the narration of problem-solving incidences where AI has been successfully applied.
Human activities are threatening the survival of millions of species on this planet. They are also accelerating the factors that contribute to species extinction, such as climate change. Some experts believe that we are in the midst of the sixth mass extinction, in some sense like the one that caused the dinosaurs’ annihilation.
One of the human activities that does just that is poaching. Rhinos, elephants and other animals are in danger of being wiped out. Conservationists believe that at the current poaching rate, elephants will be extinct within the next ten years. Despite the government’s best solutions to combat poachers, elephants’ numbers continue to dwindle.
The Mara Elephant Project attempts to solve this problem by active anti-poaching work using game rangers and research and monitoring of elephants via GPS.
To identify poachers, rangers installed a trap camera system in the Maasai Mara Reserve. This system takes photos of anything that moves, which means that alot of photos that do not involve poachers can be generated (false positives).
Going through all this photos looking for poachers is tedious. A poacher might be identified, but it might be too late.
Intel partnered with Resolve, a non-profit N.G.O that uses tech to solve real world problems. Together they came up with ‘TrailGuard’, which is a trap camera but with enhanced specs. It had a VPU chip, which used AI (object detection) to sift through images captured by the camera, and pick out only those that have captured a poacher. The camera transmits this to the MEP data centre, and rangers can be sent to that area immediately to make an arrest. Swift apprehension of the poachers, even better, red handed
while setting up snares, will save an elephant or rhino and put the culprits to justice.
As strange as it sounds, eating beef contributes to environment pollution. Cows produce methane, which is a greenhouse gas. Some researchers estimate that eating one beef burger is equivalent to driving an ICE vehicle for sixteen kilometres.
Almost a third of the world’s total farmable land is solely dedicated to rearing of animals for meat, and I think it could be put to better use. Some people have come up with very good methods to develop plant-based foods as an alternative to animal products.
One of them is Matias Muchnick, co-founder and CEO of NotCo, based in Santiago, Chile. This company, does not just develop alternatives, instead it creates products very similar to animal foods, such as milk and eggs, out of plant products only.
How is this done? The company uses an algorithm created, based on the correlation between molecular components of food and the human perception of taste, colour and texture. The AI analyses the food molecular components, creates a list of ingredients based on its most basic building blocks. Then it recombines select elements from plant foods to recreate the taste and texture of the original. It does this via machine learning and a massive dataset.
NotCo’s products are so good, that they are the third biggest mayonnaise suppliers in chile, among other products. This is despite the fact that 90% of their consumers are not vegans. They are so good that one is unable to differentiate between them and original stuff.
Can AI help us predict destructive natural events?
I refer to those which are trickier to predict, such as earthquakes.
Along geological fault lines, earthquakes are most likely to occur, but its difficult to determine when exactly. Seismic sensors can be placed within the earthquake prone area to detect vibrations, however slight they may be. These sensors generate lots of data overtime, and that includes vibrations from moving vehicles or construction equipment. To filter out these, the waveforms generated from the sensors are fed into an algorithm, which has been trained to recognize these ‘noisy’ waveforms. The algorithm analyses the data and raises an alarm if it fails to match any of the noisy data, and that most probably could be an earthquake signal.
Well, how early can an earthquake be predicted?
The average time is about a minute, which might be still too late to save lives, leave alone property. Mr. Chris Malone, a geophysics professor at Penn
State University, managed to simulate an earthquake inside a laboratory, by placing a granite block in a hydraulic press.
As the blocks are subjected under
massive pressure, measures of structural failures are recorded, until the blocks finally crack.
Machine learning is then used to find patterns in the data generated after repeating such experiments, so that a clue to an incoming earthquake can be revealed. These failures, which are similar to micro-earthquakes, are too faint for the human ear, but they are the key to predicting earthquakes, and data patterns generated from AI can help do this better.
The only challenge is replicating the lab simulation on the real world. If the parameters are changed accordingly to match real world situations, this technology
will very soon be in use.
Now that AI is good at recognizing
patterns, can it be used to prevent impending catastrophes?
Mr Mark Johnson co-founded Descartes Labs, a company which builds AI models from satellite imagery in attempt to predict calamities in key areas, such as agriculture and
renewable power generation. This can help government plan prevention or mitigation measures.
For example, planting drought resistant crops or sending
food aid in the event of a impending drought to prevent famine. The AI models are built on image recognition algorithms.
Using these, the team could be able to monitor both weather patterns and plant health. Comparing several images of
the same plantation over time, crop production could be predicted. Food insecurity
could cause unrests and political instability (like the Arab Spring). Due to this, AI based forecasts by Descartes Labs attracted the attention of DARPA, the defense
research organisation of the U.S. Military.
AI enables us to make better decisions. It might not help us prevent disasters, but with tools such as machine learning,
image recognition, predictive modelling among others, we are at least better placed at preparing ourselves better.
Credit to the Age of AI by Robert Downey Jr