Transform's great debate: AI's role in a sustainable future

On July 4th, as voters streamed to polling stations to cast a vote that would ultimately change the future of Great Britain, we at Transform were set for important conversations of our own. And so, with sustainability at its core, we set ourselves a thought experiment in the form of an hour-long debate on how AI fits into that future. With two polarised points of view, six well-researched speakers and an audience burning with questions, we came up with some compelling arguments on both sides.

Though a broad conversation, we did see some overarching themes, with concerns and optimism stemming from environmental impact, social value, ethics and innovation. To bring the conversation to you in a fully balanced way, we decided to break down the ‘for’ and ‘against’ arguments accordingly. Read on and get ready to choose your fighter.

ON ENVIRONMENTAL IMPACT

AI will never be sustainable arguments:

• Pre-AI internet data was estimated to grow at about 23% per year, however with the actual amount of computerequired to train these large language models and generative AI, we’re looking at a substantially bigger number. Though the energy efficiency of that compute is improving, the renewable energy grid might not be ready for the pressure.
• Google has recently stated a 50% increase in emissions over the past five years, primarily due to AI, which leaves them uncertain of meeting their Net-Zero targets. Microsoft have also admitted their emissions have grown by 29% since 2020 due to data centre expansion.
• AI is resource-intensive; ChatGPT uses 10x more power than a straightforward Google search and the Google AI search is 30x that, however Google are bringing the feature to a billion people by the end of 2024. In addition, AI-optimised chips and processing units uses 2x to 3x times more power.
• AI isn’t just energy-intensive, it’s also thirsty and every 10-20 ChatGPT responses evaporates 500ml of water to cool the servers instead of the half milliliter a Google search uses.
• Producing the hardware for AI requires intensive mining and mineral extraction like silicon.
• Because this is a voyage into the unknown, a lot of the time the resources required turns out to be more than forecasted.

AI can be sustainable arguments:

• AI can replace otherwise carbon intensive industries by optimising route planning to remove flights or remove the need for other oil intensive activities. With one of our own clients, we’ve optimised water networks, reducing their need to erect buildings, thereby reducing the carbon-intensive nature of constructing infrastructure.
• Though AI is resource-intensive, we can mitigate some of the effects by adopting more sustainable energy sources, using specialised chips for lower power consumption and using energy efficient algorithms where models are being compressed to reduce the size of computational requirements.
• Data centres are producing a lot of heat, but there’s an opportunity to recover it. Instead of using water to cool the centers, we could use the excess heat to warm up homes and schools.

ON SOCIAL VALUE

AI will never be sustainable arguments: 

• Students are becoming more reliant on AI software in educational settings and there are concerns around how that dependence is stunting their development in grammar, writing and critical thinking.
• With AI being able to generate art in seconds or write anything from a screenplay to a book based on a single prompt, we need to consider how that will limit the individuality and creativity of future generations.

AI can be sustainable arguments:

• Applications such as Be My Eyes show the great potential of AI to transform life for those with accessibility needs, contributing to the Sustainable Development Goals of social equity and quality of life.
• AI can also transform education by personalising the way people learn to make it more accessible. Duolingo, as an example, is a great tool for learning languages, but AI learning is of even greater advantage in areas where there is a shortage or altogether lack of teachers.
• There are countries using AI to assess tax collections and cash transfers to their benefit systems that are driving economic development and improvements in labour and productivity.
• There’s a massive need for upskilling our workforce in any event, AI has merely accelerated that; retraining programmeshave been stepped up to ensure we’re mitigating impacts of inequality.
• Only 19% of lower income countries have access to simple healthcare diagnostic methods but the use of AI can give people who’d otherwise have no access to a critical health benefit, access to it. The Lancet Commission identified 1.1 million deaths associated with the lack of diagnostics, so it can quite literally be lifesaving.

ON ETHICS

AI will never be sustainable arguments:

• Ethics and the real-world consequences that unregulated AI can have cannot be ignored. There have been cases where innocent Black men were misidentified and arrested for crimes they didn’t commit or immigrants of colour who had successfully made refugee claims, were stripped of their status due to misidentified facial recognition software. Error rates for misidentifying light-skinned men is 0.8% compared to 34.7% for darker skinned women.
• 20 to 30% of women and racial minorities are already being misdiagnosed in comparison to white men; if AI is being fed the same information from textbooks that doctors are learning from, how can it more accurately diagnose in comparison to doctors who are trained to think more holistically?

AI can be sustainable arguments:

• Ethical frameworks are being put in place to counteract inequalities; IBM, for example, has implemented their own AI toolkit with measures to reduce bias. The EU AI apps also have a whole host of regulations around AI and how it can be applied which will come into force soon.
• San Francisco is one of the first cities to ban the use of facial recognition, which shows human agency can be applied.

ON SOCIAL VALUE AND INNOVATION

AI can be sustainable arguments:

• The first public electricity generator in Britain was installed in 1881 and electric fridges weren’t invented for another 35 years, an innovation that transformed food production and some aspects of medicine as well. If we had banned electricity upfront, we’d never have known those benefits.
• We need to do whatever it takes to mitigate and manage the negatives of AI, because if we don't, we're just shutting down some incredible possibilities for it to be a force for good.

As our judges rightly pointed out, one group can point to proven negative environmental data, like carbon emissions and water usage, whilst another group had to paint a picture of the potential of a more accessible, inclusive and positive future driven by AI. At Transform, we’re contributing to that future, focussing on using AI in use cases where we can project net benefit to economy, environment and society.

So, who do you think won over our judges with their exceptional rhetorical skills? Let us know on LinkedIn and if you have any other argument we didn’t touch on, we’d like to hear your opinion at transformation@TransformUK.com.