Issues and Opportunities for AI, The Environment, and Human Rights

by Verity Anne Toews Thomson

(Images were created using AI by the author.)

The advent of artificial intelligence (AI) is bringing a paradigm shift to how and what we can use computers for. In the same way, the internet changed not just how we conducted our work, but also changed how we conducted our lives, AI will be bringing similar, and some say far greater, changes in the coming years. As someone concerned about the environment and dedicated to protecting human rights, this new tool brings with it significant opportunities to better advance this work and serious concerns about how it could perpetuate and accelerate the collapse of our planet.

What is this new tool?

For those of us that do not work in the tech industry, it may be difficult to comprehend the magnitude of the shift that AI is from our current computing. At a lecture by Google DeepMind, Drew Purves explained that the computational power in AI is 10 million times greater than our current computers. What you could ask the person who has read a single book on a given subject is relatively limited, but, if they have read 10 million books on that subject, what you can ask them and how you could use them is completely different. Simply put, we need to get used to machines doing things we did not think they could do.

Opportunities for advancing Human Rights and the environment

The implications of this new technology are far reaching. The opportunities for using AI to further environmental protection are certainly significant and are already beginning to be realized. Currently, scientists are using AI to review and curate data collected in the field. For example, scientists at the Zoological Society of London are using AI to review thousands of hours of wildlife monitoring data from camera traps. AI is also able to take unstructured data and extract value from it in a way that would take humans weeks, if not months, to do. Last year, I co-authored an article analyzing how human rights were considered in climate change adaptation plans globally. The research for this paper took four research associates several months to complete over the summer of 2020. With AI, this analysis could be run in hours. 

Advances like these can help create the research which is urgently needed to understand and respond to the climate crisis and other environmental catastrophes which pose serious risks to human rights (as more courts are increasingly acknowledging). Without understanding the effects of human action on the environment and on human rights, we cannot advocate effectively for change which protects human rights. 

Even when we know what change is needed, actualizing it can difficult. AI can assist people to quickly determine what actions are needed to decarbonize our societies. For example, one London based start-up is using AI to rapidly deliver blueprints of how best to retrofit buildings. 

AI is also able to create predictive modeling. How and to what degree this could be used in legal settings to aid in proving causation (a frequent sticking point in climate change litigation) is yet unknown. However, AI could greatly increase the availability and the complexity of predictive models to better demonstrate the effect that corporate or individual action may have on the planet.

The idea that AI will significantly democratize information is also frequently cited by its supporters. Currently, legal knowledge is largely in the hands of elites. While AI alone will not solve the problem of access to justice, it could help overcome the knowledge barrier faced by many marginalized communities. How many rights-based environmental claims have been dependent on lawyers from Western countries choosing to champion the case because local communities had no understanding or access to legal knowledge? For example, the successful class action of seaweed farmers in Indonesia was championed by the Australian Lawyers Alliance through extensive pro-bono research work before obtaining funding from Harbour Litigation Fund. With AI, anyone could ask it questions about how to formulate a claim, and even have it write many of the legal documents needed for such a case.

Issues

As much as it brings great opportunities, AI also poses potentially catastrophic problems for human rights and the environment.

AI uses considerable resources to run: every AI search uses 5 times more energy than a standard search. In Davos this year, the CEO of Open AI (maker of Chat GPT) said that AI would consume vastly more energy than expected and that an “energy breakthrough” is needed by – which he meant a breakthrough in renewable energy. The full extent of AI’s resource needs are unknown, with some estimating AI systems will use as much energy as entire nations in the coming years. 

Energy is not the only problem, water is also used to cool the AI systems and data centers. Microsoft’s water usage jumped 34% year-on-year, according to its most recent environmental disclosure, which is widely assumed to be from AI. A recent study at the University of California estimated that asking Chat GPT a string of between 20 to 50 questions will use roughly half a liter of water. The authors of the study urged further disclosure from the large tech companies to better understand the environmental impacts of generative AI. AI will also produce considerably more e-waste than standard searches, again the full extent of which is unknown. 

The human rights impacts are clear. Water, a recognized human right, is already scarce and becoming more so. In a world where between 2 to 3 billion people already experience water shortages for at least once month a year, we must think carefully about whether we want AI to consume such vast amounts of this precious resource. Similarly, e-waste poses significant human rights risks as documented by the Centre for International Environmental Law which notes that the toxic substances in e-waste products – which are informally recycled in developing countries – cause serious health impacts.

The potential problems for human rights and the environment extend beyond resource intensity. While AI will hopefully democratize information, it may also have the reverse effect: aiding in disseminating vast amounts of misinformation. The effects of social media on political polarization are already well documented. With AI, anyone can quickly produce an article on any subject, which could flood the internet with misinformation. As Michael Khoo, Climate Disinformation Program Director at Friends of the Earth, says “We can see AI fracturing the information ecosystem just as we need it to pull it back together.” This is a particular threat in the context of climate change, which is already engulfed by misinformation.  

Finally, AI has the potential to deeply entrench existing biases. The responses delivered by AI are based on its training data. That data may reflect existing human biases against minority groups ensuring that decisions made by AI are also biased. For example, a study from MIT found that AI-powered facial analysis had significantly higher error rates for minorities, and in particular minority women. It is not only the data itself, but also which data is chosen to be included in the training data which may lead to bias. The data samples may over or underrepresent certain groups, leading to bias in what AI produces. From a human rights perspective, having technology which entrenches certain types of bias is deeply concerning, particularly when this technology becomes widely used and lauded as a solution to human bias

Concluding thoughts

While technology has created and perpetuated climate change and other environmental disasters, it must now be part of the solution. AI can and must be used to help solve the climate crisis and protect human rights. Right now, with AI in its infancy, we have the opportunity to influence how it will be used in the coming decades. For those of us concerned about the environment and human rights, now is the time to engage with AI. We must question its impact and how it works, advocate for regulation, and ensure that we are involved in its implementation and development.



Posted

in

by

Tags: