|
Getting your Trinity Audio player ready...
|
Artificial intelligence has a presented a variety of speculative problems, from job automation, to an excuse for employers to weaken worker power by depressing wages and working conditions, to apocalyptic scenarios where AI takes actions against human life.
What often goes unsaid are the tremendous environmental costs of AI, which are hidden from the average consumer. Whether using predictive AI – statistical analysis with machine learning algorithms to find data patterns and forecast future outcomes – or the more energy-hungry, generative large-language models (e.g. ChatGPT, Claude, Perplexity), we only see text and images appear on a screen – but the responses are fuelled by a tremendous network of data centres.
A 2024 report from the UN’s Environment Programme summarizes environmental concerns around the data centres powering AI, including:
- The production of computers for data centres is materially intensive: Making a 2 kg computer requires 800 kg of rare-earth minerals that are mined in unsustainable ways.
- Data centres create hazardous materials including mercury and lead.
- Data centres use extensive amounts of water to cool equipment that powers AI. (The combined global water use of Google, Microsoft and Meta is predicted by 2027 to be equivalent to the annual water withdrawal of 4-6 Denmarks.)
- Energy consumption of generative language models is much higher than traditional models, with a request to ChatGPT consuming nearly 10 times as much energy as a Google search.
Perhaps most concerning of all is that these environmental resources and energy have been used by large language models to generate environmental misinformation that can delay the need for environmental regulations and action necessary to stop our present climate crisis. The effects of the climate crisis are all too evident and require regulations that reflect the true cost of AI – not only in dollars and cents, and work hours and job security, but in allowing our world to be habitable for all of us to enjoy.
We must ask ourselves, is the product of AI worth the cost, economically and environmentally? If not, what can we – as Canadian career professionals – do?
Press for regulations
One challenge is the absence of a regulatory framework in Canada specific to AI. There are several major signatories to Canada’s Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, but it has not been adopted by major AI companies including Google, Meta or Microsoft. The Canadian Artificial Intelligence and Data Act (AIDA), which aims to regulate business use of AI and prevent discrimination, has yet to be enacted. If this matters to you, let your Member of Parliament know that this is a voting issue!
We must ask ourselves, is the product of AI worth the cost, economically and environmentally?
Europe offers progressive examples of environmental and human rights regulatory frameworks in the tech sector. Starting in 2024, the International Energy Agency placed mandatory reporting obligations for the energy use and emissions from data centres, and set climate neutrality targets.
The Germany Supply Chain Law goes further, requiring that companies with over 1,000 employees must ensure environmental and human rights standards are met throughout their supply chains. Companies that fail to respect the terms of the law can face fines of up to 2% of the company’s annual revenues.
Leverage the power of organizing
Beyond political regulations, there is the power of collective action. Author Brian Merchant writes that blaming machines for job loss masks the agency companies have in deciding to automate jobs. Full automation is rare right now, but it is being used to degrade working conditions and depress wages. Arguments that automation is inevitable leads to a dearth in critical thinking about where automation is best implemented.
However, some workers are pushing back. Recent U.S examples include the Hollywood writers’ strike of 2023, Tech Workers Coalition’s climate strikes and opposition to use of AI by immigration, military and police agencies. These examples highlight the power of collective social refusal. As a career professional, if you are working in a unionized environment, you can insist to your union representative that AI regulation matters to you. The same advice can apply to clients and students who are engaged in this issue.
Share resources
Finally, it is important to be aware of tools that can help our clients prevent AI from using stolen work to depress wages. A small example is that artists be made aware of programs such as Nightshade and Glaze to disrupt AI learning algorithms from extracting content from artists without their consent. Another is Nephentes, which traps web crawlers scraping website content without permission for large language models.
All this matters because behind AI’s glossy exterior is a swath of environmental damage and ultimately the loss of jobs for workers. We use terms such as transition and reskilling as if they are inevitable and that everyone will make the transition. Writers Angus Case and Anne Deaton remind us that “jobs are not just the source of money; they are the basis for the rituals, customs, and routines of working-class life […] It is the loss of meaning, of dignity, of pride, and of self-respect.”
It is important that we understand the human and environmental cost of AI, lest our careers and the planet are consumed by it.
