The Advance of AI
8 minute read
The twelfth article of Earth Matters concludes the series with a more futuristic vision of sustainability powered by Artificial Intelligence (AI). Advances in areas like agriculture [1], Big Data, resource management and conservation [2] already reveal AI to be a powerful ally… But there is still an environmental cost to be considered. In this article we explore both the good and the bad, and look at AI’s role in sustainability across three dimensions: how it can augment our abilities as protectors of Mother Earth; how it can act as a corrector of human error; how its capacity to act as a smart predictor for the future empowers us to make more efficient (and sustainable) decisions.
* * *
Throughout 2020, Earth Matters has been our monthly thought piece that followed the path to sustainability, helping us to find tangible progress for people and planet.
This twelfth article is the last of the series, which covered a wide range of topics, from rethinking luxury to reimagining waste, from becoming sustainable across all generations to moving societies towards a Circular Economy.
Earth Matters will continue to evolve in 2021, as we revisit the twelve topics with fresh eyes, ideas and conversations, keeping this important subject alive, moving and open. Stay tuned!
***
AI trailblazer Marvin Minsky once defined AI as “the science of making machines do things that would require intelligence if done by men.”[3] Indeed, the idea of computers thinking more (or even better) than humans has been discussed and debated by scientists and researchers since the pivotal work of Alan Turing in the 1950s.[4]. However, this broad definition has lead to many debates over what makes something truly AI, and the associated implications for sustainable innovations.
A measure of intelligence.
Acknowledging that AI is still relatively in its infancy, researchers have defined three stages of intelligence, going from weak to conscious, as summarised below.[5]
ANI is now ubiquitous in our day to day lives. These are the programs and algorithms that enable Facebook to automatically recognise faces when tagging photographs, that run Tesla’s autonomous vehicles, and that allow Siri to respond to our voices. AGI and ASI are evolving rapidly and in the future it is predicted that we will see more AI that is able to reason, plan, perform tasks and find solutions to problems beyond what it was initially designed for, even surpassing human capabilities.
The good…
The potential of AI and the possible benefits to our environmental and social challenges are one of the concepts being explored by the World Economic Forum. In a 2018 report (developed with PwC and the Stanford Woods Institute for the Environment) they highlight eight AI “game-changer” applications that could deliver truly transformative solutions for people, planet and profit.
They also identify six priority action areas where AI could help address the planet’s challenges.[6] These are initial areas for sustainable, AI-powered innovation, all stemming from the fact that “we stand on the brink of a technological revolution that will fundamentally alter the way we live, work, and relate to one another.”[7]
Priority action areas for addressing Earth challenge areas, Harnessing Artificial Intelligence for the Earth
The processing power, storage and access to data from the billions of people with interconnected devices, combined with AI and emerging tech – such as robotics, the Internet of Things, autonomous vehicles, 3D printing, nanotechnology, biotechnology, materials science, energy storage, and quantum computing – could lead to sustainable end-to-end solutions.
Taking agriculture as an example, in the future we could see autonomous farming systems with farmers growing crops symbiotically, using AI to forecast problems and automatically take corrective measures using robotics. Crops could be harvested and processed in highly efficient lights-out facilities, delivered using autonomous vehicles, running on green energy, and with logistics optimised for efficiency using AI powered predictions.
“Put simply, AI can enable our future systems to be more productive for the economy and for nature.”
– Celine Herweijer, Global Innovation and Sustainability Leader, PwC UK
…And the bad.
However, things aren’t all rosy in the AI camp… Scratch beneath the surface, and the sustainability story becomes a little murkier.
There are many debates about the ethical and social considerations still surrounding AI, such as job displacement, AI biases, transparency of algorithms and conformation to ethical principles (particularly vital in industries such as medicine and finance).[8]
Another problem area of AI that is often overlooked is the environmental footprint of developing, teaching and running the systems. For example, a study released in 2019 by MIT Technology Review [9] found that training a standard AI has the same carbon footprint as a flight across the United States. And training a more sophisticated AI is even worse, with over five times the carbon footprint than the entire lifecycle of an American car, including manufacturing.[10]
AI learning, algorithms, active systems, or even streaming on TikTok – these activities generate and consume data. This data must be captured, stored, analysed and distributed, all of which requires huge amounts of processing power. The amount of data that we generate, consume and send is so enormous it is almost incomprehensible. One estimate suggests that over 33 zettabytes (ZB) of data were floating around the world in 2020. By 2025, it is predicted that this number will grow to over 175 ZB… That’s 175 trillion gigabytes![11]
The challenge of keeping data centres cool is becoming increasingly difficult, with machine learning contributing to this problematic heat density. Combined with more advanced knowledge of the carbon footprint of these data centres, the tech industry is being forced to innovate and think more strategically about how they deal with data. Some innovative solutions include using geography to aid natural cooling, positioning centres in countries with cooler climates.[12] Other solutions look at using more efficient cooling technologies such as liquid cooled centres.
This Google Data Centre in Dublin takes advantage of Ireland’s temperate climate to efficiently cool the servers.
Google have taken an even more innovative approach, with their DeepMind division developing AI that teaches itself to minimize the use of energy to cool Google’s data centres, enabling Google to reduce its data centre energy bill by 40%.[13]
“Throughout history, every generation has confronted big challenges. Climate change will be our generation’s most profound challenge – and in 2021, the world will take its biggest steps yet to meet it.”
– Sundar Pichai, CEO Alphabet Inc
PwC predicts that by 2030 quantum and distributed computing could be readily available, which would mean that instead of the inefficient brute force methods currently used to increase the computing power of AI, quantum computing would open the door to highly efficient machine learning.
However, with quantum computers currently needing temperatures close to absolute zero to operate (that’s colder than the vacuum of space),[14] this raises the question: could AI end up being used to solve its own environmental shortcomings?
Protector
In the above-mentioned 2018 report, the World Economic Forum highlights several potential areas where AI might work as a protector for sustainability in all dimensions, including habitat conservation and restoration, sustainable fishing, and preventing pollution.
AI has also played a key protector role during the Covid-19 pandemic. Social distancing, increased hygiene and safety measures, remote working, and a greater reliance on digital connection have resulted in an accelerated adoption of AI, automated systems and robotics.
So what are some examples of brands and organisations harnessing the protective power of AI for people, planet and profit?
Robo cop
Silicon Valley based tech company Knightscope has developed Autonomous Security Robots (ASRs) that have proven invaluable during the pandemic to guard empty real estate. The patrol robots deliver real-time intelligence, feeding back information and data to human security teams, thereby enabling them to take fast action, increasing safety, reducing risk of infection, and acting for social responsibility.
Listen and learn
The Zoological Society of London joined forces with Google Cloud to protect endangered species. They installed 69 acoustic recording devices in northern Cameroon, capturing 267 days’ worth of audio (and resulting in 350GB of data). Google Cloud’s AI and machine learning tech then identified and labelled the sounds (including gunshots), helping to identify and track endangered species in days instead of months.
Super sorter
Finnish recycling firm ZenRobotics has enhanced the recycling process by creating a fast picker for lightweight materials (including packaging). Through machine learning, the robot learns from information provided via NIR spectroscopic cameras, 3D laser scanners and metal sensors installed along the conveyor belt, and it makes its own decisions on what materials to pick and sort, all without human intervention.
Corrector
Developments in AI, robotics and smart automated tech create exciting cross-industry opportunities, from healthcare to farming to manufacturing. The ability of automated, AI driven robotics to take over and flawlessly perform repetitive tasks means increased efficiency, elimination of human error and waste reduction, benefiting both the environment and companies’ bottom lines.
Smooth sailing
Tech company Semcon and marine coating company Jotun have created an underwater smart cleaning robot that reduces cargo ship emissions. When a ship stops at port, the HullSkater robot travels over the hull, removing the build-up of algae and microorganisms and improving hydrodynamics. The HullSkater cuts emissions by 12.5% and helps to reduce the spread of invasive species from one port to another.
Record breaker
Nike has invested in Grabit robots, which can precisely assemble all the pieces needed to make the most labour intensive part of a sneaker – the upper – in 50-75 seconds. A human worker typically takes over 10 minutes to do the same job. Grabit uses electroadhesion (i.e. static electricity) to quickly pick up and handle varying materials. This results in faster assembly times and a reduction in errors.
Intelligent move
A UK start-up has created the first autonomous driving system that learns from feedback. Wayve’s machine-learning platform is unique in that it can drive on roads not seen during training without a HD map of its environment. Each time a safety driver intervenes, the Wayve car drives better, learning from experience, example and feedback. This paves the way for more efficient, safer and sustainable travel.
Predictor
We are facing a climate crisis, with global heating leading to a number of natural disasters, including floods and droughts (as covered in our previous article, A Watertight World). In India, Google is developing an AI platform that warns users of impending floods through Google Maps and Google Search.[16] In addition to flood prediction, AI and satellite imaging is being used to predict and issue early warnings of other natural disasters, from volcano eruptions to hurricanes, earthquakes and tsunamis.[17]
AI is also being used as a smart predictor across many industries, from farming to logistics to fashion. Its foresight enables brands and companies to manage more efficient supply chains, and these gains will continue to grow as AI becomes smarter and more interconnected.
Farmers forewarned
Founded in Israel in 2017, tech company AgroScout uses AI for the early detection of crop disease, enabling farmers to take pro-active measures. Leaf level visual detection systems identify bugs and diseases, while drone and smartphone software provide real time feedback on crop health to farmers. The tech allows farmers to address problems early on, improving yields, reducing pesticide use, and increasing profits.
Smart and in-stock
Indonesia-based Sorabel leverages AI to maintain a low inventory risk while promptly delivering fashionable items at affordable prices. The AI powered system predicts fashion trends and provides forecasts on the styles that are most likely to sell. This cuts waste by producing only items that customers are likely to purchase, preventing excess stock and PR catastrophes such as the Burberry burning scandal in 2018.[18]
Drone deliveries
To deal with the unprecedented demand for postal services due to Covid-19, FedEx used robotic arms, smart cameras and Bluetooth-linked pods to manage 2020 Christmas deliveries. They are also testing aerial drone deliveries from Wing, a division of Alphabet (Google’s parent company). Automating deliveries and using the power of AI can increase efficiencies and support in typically troublesome “last mile” logistics.
François Chollet – a researcher at Google and creator of AI machine-learning software – simplifies and updates today’s definition of AI, describing it as a system’s ability to adapt and improvise in a new environment, to generalise its knowledge and apply it to unfamiliar scenarios. [19]
This and the examples and ideas explored so far all point to the same conclusion: the true capability and magnitude of machine intelligence are still very much to be discovered. And there is an ever-widening scope for AI to help us perform better and more sustainably. However, human passion, commitment and ingenuity will continue to be fundamental to finding ways to ensure that AI can be used safely, ethically and responsibly, for people, planet and profit.