AI: what is the social and environmental impact?

Concerns about the use – and misuse – of AI encompass employment, the environment and ethics. Will it lead to job losses? Does it use too much of our scarce energy and water? And what is the risk that it is used to spread misinformation?

Drawing on research from Schroders sustainable investment specialists, we look at AI through the lens of the environment, human rights and human capital management.

Environment

AI is often considered to have minimal physical impact because it operates in the digital realm and exists as lines of code and algorithms. In reality, it is grounded and enabled by very real, physical infrastructure – such as datacentres, processors and other specialised computing hardware – all of which have substantial raw material and natural resource requirements.

Let’s start with one of the biggest direct beneficiaries of AI today, the designer – and indeed inventor – of the GPU (graphics processing unit), Nvidia. GPUs are small but powerful devices that facilitate machine learning, video editing and gaming. Building a chip is one of the hardest manufacturing processes in town. Silicon wafers are sliced, diced, layered and assembled with some of the most precise equipment on earth. A single speck of dust can compromise their functionality.

Nvidia, of course, only designs the chips, outsourcing manufacturing to other important companies, such as TSMC. This means that the manufacturer’s scope 1 and 2 emissions (those directly owned or controlled by the company) are higher than Nvidia’s. But the carbon has been emitted as a result of bringing Nvidia’s innovation to life. We therefore need to consider the environmental impacts that occur higher up in the supply chain, tracing back to the extraction of raw materials, as well as down to the end of its useful life. While complex and notoriously opaque, these scope 3 emissions (those arising from the production, use and disposal of the products) are currently the best estimation investors have for measuring the real-world emissions impact across the lifecycle of a product.

The same thought process should be applied to water use. Technology is a thirty sector. Indeed, a recent study by the University of Massachusetts Amherst found that training a single generative AI model can consume as much as 284,000 litres of water, equivalent to the amount of water an average person would consume over the course of 27 years.

More positively, AI could help us address some of the world’s most pressing environmental challenges. By enabling better monitoring of resources, it could help us make more efficient use of energy and water. AI-enabled “agritech” could help improve food security while it could also improve our ability to predict extreme weather. 

Did you know?

• AI is a digital technology but it depends on physical processes with real environmental impact. Datacentres consume huge amounts of energy and water and making semiconductors requires incredibly sophisticated equipment and facilities.
• AI could also help us solve climate challenges. Nvidia’s technology is now being used to create a “digital twin” of the earth, allowing for climate change modelling and scenario analysis.
• Find out more here.

Human rights

Digital access has made a huge contribution to economic development. While the percentage of the global population with access to the internet now exceeds 60%, billions of people remain without. As technology evolves, this “digital divide” could continue to prevent equal participation for low-income households, the disabled, rural areas, and older adults.

Against this backdrop, there are concerns about the deployment of generative AI while many still to do not have basic internet or broadband access. The UN has called for a Global Digital Compact (to be agreed in 2024) to promote increased transparency and accountability about the impact of digital technologies on human rights.

As with the environment, we also need to consider the benefits that AI can bring to human rights. Studies have already shown how AI can improve outcomes in education and healthcare. These exciting initiatives will stand on firmer ground if good governance practices are adhered throughout the AI lifecycle. Firms interacting with AI should consider the completeness and traceability of data going into the system, as well as the fairness and accuracy of what comes out of it.

Did you know?

• AI could help teachers reallocate 20-30 percent of their time towards activities that support student learning, according to recent research by McKinsey. Automated marking and reducing the burden of clerical work are potential starting points.
• AI can improve the speed and accuracy of medical diagnoses and reduce the subjectivity of decision-making. However, practitioners are warning about “algorithmic bias,” exacerbating existing inequalities. For example, if an algorithm is trained with a dataset overrepresented by male patients, it is possible that women are more likely to be misdiagnosed.

Human capital management

We’ve been automating work for 200 years. Every time we go through a wave of automation, whole classes of tasks go away, but new classes of task, and new jobs, are created. There can, of course, be dislocation in the process and sometimes the new jobs go to different people in different places. Over time, however, the total number of jobs has not gone down, and society, generally, has become more prosperous. We engage with companies we invest in on their approach to human capital management and expect them to act responsibly in their efforts to attract, retain, develop and motivate their employee base.

Most jobs today involve the assimilation of a wide range of skills and attributes. In the context of financial advice, for instance, AI could probably tell an individual how much they should be allocating to a pension – but does it have the empathy to understand a client’s personal circumstances, and can it earn their trust?

For a workforce to make the most of new technology, a certain level of upskilling and reskilling is required. Today, technological and innovation capabilities are highly sought-after skills for hiring companies, but they sit amongst a wider and more established wishlist of very human skills spanning leadership, cognitive reasoning and creative problem solving.

AI may create the opportunity to automate certain tasks in a way that improves speed or accuracy. This may result in workers having more time to do focus on the human-centric elements of their job. 

Managing knowledge in a world powered by AI means embracing the symbiotic relationship between humans and AI. AI can filter, sort and process high volumes of data and discover unnoticed patterns, while humans can train and critically assess intelligence assistants, develop tacit knowledge through social interactions that relate to AI, and, perhaps most importantly, take responsibility for the outcomes and recommendations that arise from its use.

Did you know?

• New technologies raise productivity. One indication of this is the fact that the average number of hours worked per person per year has been declining since the late 1980s, despite new jobs being created. Today, the average worker in developed economies works for around 8% fewer hours than in 1987, according to the OECD.
• Economies have already been through many huge shifts in employment and seen their workforces grow and become wealthier. In 1870, over 50% of US workers were employed in agriculture, but by the end of the 1920s the figure was just over 20%. Today, the figure is roughly 1%.