A drone offers a unique bird’s eye view as part of a forest fire surveillance system in the Landes region, France. Photo by Pierre Andrieu/AFP/Getty


Algorithmic wilderness

Robo-bees and drone-seeded forests: can technology mend our broken relationship with the natural world?

by Henry Mance + BIO

A drone offers a unique bird’s eye view as part of a forest fire surveillance system in the Landes region, France. Photo by Pierre Andrieu/AFP/Getty

I’ve tried many ways to free my brain from my iPhone. I’ve invented rules, bought books, deleted apps, installed other apps. But the only thing that reliably works is to leave the phone at home and to walk along a path through the nearby woods. With trees overhead, and mud below, you quickly forget the last social media notification. You escape the internet of things by surrounding yourself with things that can never be plugged in.

Nature and technology seem to pull us in different directions. Silicon Valley prophets talk of colonising new planets, conveniently sidestepping the challenge of saving our own. Smartphones drag our attention into a vortex of high-definition screens and endless notifications, while the demand for ‘innovation’ drives an endless cycle of consumption and waste. Amazon and Twitter take their names from the natural world, even as they invent products that seem bent on supplanting it. At Google and Facebook, the world’s best engineers are engaged in souping up our own species via artificial intelligence (AI), rather than protecting millions of others.

Behind all this tension is a cultural clash between the optimism of the tech industry and the pessimism of much of the green movement. In its crudest form, conservation tends to be about keeping people ‘out’ to minimise humanity’s destructive impact; it rejects the idea that we can improve places through our presence and inventions. Conservationists are disposed to have an inbuilt skepticism of grand projects – derived partly from bitter experience, such as Australia’s notorious decision to introduce cane toads in the 1930s to eat agricultural pests (now it’s the toads who have become the pests).

Yet the tech community has always had its hippies and its environmentalists. What if technologies were to become part of the natural world, woven more closely into its rhythms and processes? You might have heard of smart cities; now we have the conceptual seeds of smart countryside, forests and wetlands too. Picture green landscapes populated by an army of self-guided bots, responding to the shifting needs of the environment in which they’re embedded, and evolving alongside it, too. After all, the most recent wave of innovation has been driven by ‘machine learning’ or ‘deep learning’ – self-teaching algorithms that get better and better at detecting and predicting patterns in data, a computer-science version of an adaptive process.

Drawing a bright line between humans and nature has always been tricky. It’s all the more difficult now, when no ecosystem on Earth is free from human influence; in the Anthropocene, nowhere is truly pristine or wild. ‘The paradox, in a nutshell, is this,’ writes the journalist Oliver Morton in The Planet Remade (2015), ‘humans are grown so powerful that they have become a force of nature – and forces of nature are those things which, by definition, are beyond the power of humans to control.’

Perhaps the answer is not to try vainly to remove ourselves from the equation. Instead, we might create a different category of place, one shaped not just by humans and nature, but by thinking machines as well – what the philosopher Huw Price at Cambridge University has described as the ‘Machinocene’. But will this new era remove a source of wonder from our world, or create a whole new one?

To start to understand the dilemma, take Mallorca, the Spanish island where tourism has left a deep footprint in recent decades. Last year, I visited S’Albufera national park, one of the richest ecosystems in the area, which my Lonely Planet guide optimistically describes as having the best birdwatching in the Mediterranean.

The park’s entrance is located halfway down a road lined with luxury hotels and shops selling rubber inflatables. The land was farmed until the 19th century; even as a national park, it requires invasive infrastructure – cement paths, a visitor centre, bird-watching huts. Planes fly overhead. In other words, it’s a human-shaped environment masquerading as a natural one. ‘If you don’t explain to people that it’s human-made, they don’t realise,’ Maties Rebassa, the park’s director, told me. ‘I always say, it’s artificial, but very naturalised.’

We’ve damaged so much that, if we want to experience something wild, it might have to be constructed

The real problem lies beneath the surface. Over the past 15 years or so, S’Albufera has become overrun by carp. The park’s authorities are unsure how the population started, but their favoured hypothesis is that someone released the fish in the hope of later fishing them. You can now see the glistening bodies of the creatures, bobbing under the surface. By displacing sediment in the canals, the carps turn the water muddy, and prevent plant life from growing.

Each year, using nets, hooks and cages, the authorities extract hundreds of thousands of carp from the waterways. But the younger, smaller fish elude capture, and there are huge areas of reedbeds that the park officials can’t even reach. The result is a losing battle. The authorities can’t drain the park, because it’s connected to the sea. They’ve even considered using genetic alteration of the carp to somehow bring about a population collapse.

For people fighting such intractable environmental wars as this one, the appeal of new ‘smart’ technologies is clear. ‘The idea that autonomous systems could solve complex problems about conservation is inevitable,’ says James Canton, a self-described futurist who has held appointments at the Massachusetts Institute of Technology and at Ray Kurzweil’s Singularity University. As in S’Albufera, the reality is that we’ve damaged so many ecosystems that, if we want to experience something wild, much of the time it might have to be constructed.

‘A million dollars isn’t cool. You know what’s cool? A billion dollars.’ So claims the character of the real-life Facebook investor Sean Parker in the film The Social Network (2010). Lauren Fletcher, a former NASA engineer, has imbibed this calculus, but with a twist. ‘We’re going to plant 1 billion trees a year,’ he says in a promotional video for his business, which shows him tramping through the countryside in an anorak and glasses, before putting a plastic sheath on a seedling. ‘We’re going to change the world 1 billion trees at a time.’

BioCarbon Engineering, Fletcher’s company, with offices in Oxford, Dublin and Sydney, is banking on using drones to reforest large expanses of the Earth’s surface. Like many tech ideas, it has a beautiful simplicity. Planting by hand is an expensive process. So is planting by planes and helicopters, which has been done for decades. But drones could do the same thing in less time and for little cost. In places such as the Amazon and Borneo, where tropical forests have turned into dull pasture and plantations, mass plantation could restore some of their lost wonder.

BioCarbon Engineering’s drones scan the landscape for suitable soil, then fire out seedpods with sufficient force for them to break the surface. The company claims that it has already planted 25,000 trees in trials worldwide, and is working on replanting mangroves in Burma. A next step might be to use its drones to replenish areas hit by mining and forestry.

The problem is not reforestation – it’s preventing forests from being destroyed for agriculture and forestry in the first place

Despite its early promise, question marks hang over the venture. The biggest is whether it’s really ambitious enough. On its own, aerial planting seems best-suited to pioneer species – those trees that grow well on bare land – and even then, once seedlings are established, they generally need protection. What’s more, the fundamental worldwide problem with tackling climate change is not really reforestation – it’s preventing forests from being destroyed for agriculture and forestry in the first place. ‘Planting seeds isn’t going to solve that,’ says Glyn Davies, senior advisor at the World Wildlife Fund Malaysia.

Still, these experiments suggest that conservation is not the amateur, artisan pursuit some critics might imagine. Environmental scientists also use technology such as lidar – a form of laser surveying, similar to radar – to map forests and the threats to them. Lidar not only measures the height of the forest; it can also be used to gauge how much light penetrates the canopy, and what species might thrive below. Conservationists also deploy sensors and camera traps to detect animal movements, augmented by software that can, for example, help to identify tigers by building a three-dimensional model of their skin and stripes (no two tigers have exactly the same pattern). And machine learning has accelerated these developments. Researchers at Georgia Tech, IBM and the Smithsonian Institution are working on tools that can accurately model what will happen to whole ecosystems in response to environmental shocks – an exercise that requires AI capable of crunching decades-worth of data, from the specific mass of microorganisms to the growth processes of coral reefs.

These are examples of conservationists exploiting technology to guide decision-making. Even people who might see nature in spiritual or semi-spiritual terms – to be kept as far away from human influence as possible – probably wouldn’t object. But we should also prepare for technology to go further: to start making choices on its own, with less and less intervention from human reviewers.

For years, scientists have tracked the alarming and mysterious decline in bees around the world. As pollinators, bees play a key role in helping cultivated and wild plants to reproduce. In the United States, wild bee numbers fell nearly a quarter from 2008 to 2013, according to one model; populations elsewhere are also under threat. While European regulators have responded by banning certain pesticides, the full causes of the bees’ decline isn’t understood.

Meanwhile, researchers at Harvard and Cornell universities have been developing a potential replacement: RoboBees, which they call ‘autonomous flying microrobots’. The mini-bots each weigh less than a 10th of a gram, have sensors that mimic a bee’s eyes and antennae, and are designed with the intention that they will coordinate as a ‘colony’. Separately, Japanese researchers have tested a sticky gel that, when applied to a small robot, can pick up pollen.

However, we should beware of the seductive and misleading idea that these inventions represent a quick techno-fix. The Harvard-Cornell team hasn’t worked out how to coordinate the mini-robots to swarm like bees, or to equip them with sufficient power to handle high winds. The RoboBees are often mentioned as potential pollinators, but they might be better suited to simpler (if more sinister) tasks, such as surveillance. In any case, bees are not just pollinators – they have evolved to be part of their ecosystems in other ways that RoboBees will struggle to replicate. Bees’ bodies biodegrade. They provide food for other organisms. They almost certainly have interactions of which we are not aware. For the moment, it’s hard to imagine RoboBees replicating these functions in the short term. One commentator has suggested that drones would be most useful in environments that aren’t adapted for natural pollinators, such as ‘a biodome on Mars’ (a scenario that raises more challenges).

Pest control presents a more immediate opportunity for robotics. In Australia, outbreaks of crown-of-thorns starfish have become one of the biggest threats to the Great Barrier Reef. An individual can grow to nearly a metre across, and eat up to 10 square metres of coral a year. In some areas, the creatures are estimated to be responsible for nearly half the decline in corals in the past 30 years. Solutions from building fences to divers injecting chemicals have been tried, but none has scaled well – ‘They just don’t have enough people,’ noted Matthew Dunbabin, a robotics researcher at the Queensland University of Technology.

Natural predators to control invasive species often produce disastrous, unforeseen results

Dunbabin designed an alternative: a robot capable of identifying the crown-of-thorns starfish and injecting it with fatal poison. He emphasised the device’s advantages over human divers: it was able to stay under almost three times longer, he said, and could work day and night, though the decision about whether to inject would still require the approval of a human operator onshore. In 2016 after a public vote, Dunbabin was awarded AU $750,000 (about US $600,000) by the Google Impact Challenge, a philanthropic initiative, to develop an adapted device – a low-cost ‘robo reef protector’. If successful, such devices are likely to be given greater autonomy: perhaps the freedom to strike without human approval, or to roam semi-wild. (This is, of course, a small commitment from Google: the company currently generates $600,000 in profit in less than half an hour.)

But recall the morality tale of the cane toad. Strategies based on introducing natural predators to control invasive species have often produced disastrous, unforeseen results – and it remains unclear whether devices such as RoboBees and the robo-reef protector can avoid the same fate. Autonomous conservation systems might be ultimately subject to the control of humans, in a way that Australia’s cane toads were not, but machine learning can still go awry. Systems must self-teach in order to adapt and respond, but quite how they do so is often impenetrable, even to their creators. In the long run, such systems might even develop their own goals, or seek to alter the environment for their own benefit, rather than for other species. And they could misjudge how nature works, and make damaging alterations, which humans do not appreciate until too late.

Let’s assume that the systems can overcome these doubts – and that humans can deploy robots that replicate the role of particular species on a large scale. Even then, we might still prefer to walk through a forest in the knowledge that it has grown naturally (rather than been kickstarted by drones). Yet surely any forest would be preferable to no forest. In a drone-seeded forest, or a RoboBee-pollinated meadow for that matter, at least we’ll be able to see and appreciate the trees and the birds and flowers. Perhaps we might learn to accept the absence of ‘real’ bees, given the presence of the rest of the field.

But what if humans took a further leap, and developed self-regulating, artificial systems for tweaking and optimising and governing the natural processes of ecology? In 2017, scholars from Harvard University and the University of Maryland explored this idea, which they called the ‘automated curation of wild places’ and a kind of ‘designed autonomy’. The idea would be to use devices as ‘wildness creators’, cutting out destructive, top-down human intelligence from the picture. These technological scaffolds would apply computational analysis to nature, determine the correct conservation goals, and pursue them, working with data acquired from the bottom up. The mode is one of the ‘machine as a gardener’, to use the authors’ metaphor.

Admittedly, such green-fingered super-algorithms are not imminent. Erle C Ellis, one of the paper’s authors, told me that the idea is ‘experimental’, and it’s only likely to be tangible over the course of ‘this century’. Yet the implications are so vast that they are worth considering now, especially as ecological catastrophe closes in. The writers’ choice of term ‘wildness’ is telling. The more common term is ‘wilderness’, a place untouched by human intervention. But wilderness no longer exists. Humans have irrevocably changed the climate, acidified the oceans, and altered the conditions of life for almost every species on the planet.

The writer Robert Macfarlane found this out for himself when he began his search for Britain’s remaining wilderness. In The Wild Places (2007), he begins looking for spaces free from human direction, but comes across boats, rusting tractors and plastic debris. He decides that the human and the natural cannot be divided. So he narrows his ambition to finding wildness, and eventually argues that this can be found even in mundane landscapes, under the cover of darkness.

Wildness appears to be a state of mind, to which a sense of awe and astonishment is central. Such feelings seem to be attainable in landscapes that have been transformed by humans, so long as the influence of other natural processes are evident. By being outside our control, natural forces remind us that our own existence is transitory and insignificant – the essence of the idea of the sublime, and an experience that many people think is worth cherishing, preserving and promoting.

There’s a strain of environmentalism that pushes back against all this: arguing, in a nutshell, that fevered technological development got us into the mess and won’t get us out. Natural processes are so complex that it’s unclear how we could ever replace them. Our ecosystems are made up of species that have co-evolved and interact in ways we don’t understand. Best to leave nature alone. Rewilding, an increasingly influential approach, is the notion that humans should stand back to allow other species to flourish – an abandonment of the Biblical doctrine of dominion over the natural world, as the writer and activist George Monbiot wrote.

In Confessions of a Recovering Environmentalist (2017), the writer Paul Kingsnorth lays the blame at the door of the Enlightenment, planting the seed of humanity’s ambition to refashion the planet in its own image. ‘All we can do is argue for more of the same: more onward momentum, more technological mediation, more control,’ he writes.

Instead, Kingsnorth calls for us to embrace ‘the slow, messy business of getting to know a landscape’. This knowledge comes not through algorithms, but through feeling the earth under one’s fingertips. (In his book, Kingsnorth writes about finding particular solace in constructing a compost toilet.) It’s only deep contact with nature that will shift our thinking, he believes, and give us an appreciation of the limits that our natural world places upon us.

Ellis, one of the proponents of curated wildness, is aware of its pitfalls. He draws a cautionary comparison with the 19th and early 20th centuries, when conservationists tried to exclude local people from national parks and thereby created a backlash. It’s not a good idea, he told me, ‘for people to feel less connected or disconnected from the effort to conserve the natural world. I don’t think it’s good to create a feeling of “We could just leave it to the robots”.’

Our experience of nature could be both more technological and more fulfilling

If we follow this route, we end up with a dim view of robots and AI. However ingenious researchers might be, ultimately they’re adding another intermediary between humans and the natural world. They’re gambling that our interventions won’t cause the unforeseen cascades that they have in the past.

However, it’s also possible to glimpse a contrary perspective. Past attempts to remake ecosystems have fallen short because of limited information and top-down planning; past technologies have rarely prioritised genuine conservation. A ‘wildness creator’, by contrast, could have broad goals and long-term horizons. Robotics, sensors and machine learning could adaptively enhance natural processes much better than we’ve been able to, at least up to now.

After all, it is hardly unprecedented for organisms to shape their environment: Charles Darwin noted how earthworms burrow through the soil to make it liveable. Humans have constructed their own niche through the breeding of animal species. Machine learning is simply a shift in how the process works. In some areas, they might even allow us to gain a new awe for natural processes. We might design automated systems to eliminate noise, light and other pollution, so non-human species can thrive. We could use sensors and communication devices to warn animals of poachers or settlements, allowing humans and wild animals to share the same territory. Our experience of nature could be both more technological and more fulfilling.

We once aspired to maintain wild places free from human footprints. In many cases, that’s simply no longer tenable. In S’Albufera, there’s no option to stand back – to rewild by inaction. We need to use the best tools at our disposal to address the invasion of carp. We can reject technology, and hope that the tech optimists focus their energies on Mars. Or we can embrace it, and usher in a new era in our relationship with nature. Machines will shape our relationship with each other over the coming decades. Why should we expect they won’t shape our relationship with nature too?