By Rosemary Thomas, Senior Technical Researcher, AI Labs, Version 1
Artificial Intelligence is a transformative force that is reshaping our daily lives. It serves as an instrument of change, driving innovation across various sectors by automating tasks, providing insightful data analysis, and enabling new forms of interaction. AI is fostering a new era of efficiency, productivity, and creativity.
More importantly though, through transparency, ethical AI practices, and healthy privacy safeguards, AI can help to strengthen our trust in technology and its role in our daily lives. It is a catalyst for changing how society perceives sustainability, helping us predict and work towards a more sustainable, ethical future.
Making a difference with AI for good
‘AI for good’ pertains to the use of AI technologies to help solve specific societal challenges and contribute towards making people’s lives better. It leverages the strength of AI to address issues like economic hardship, physical and mental wellbeing, academic achievement, and the preservation of nature.
For businesses, ‘AI for good’ can mean using AI to contribute towards environmental, social, and governance (ESG). Used correctly, AI can help to create sustainable strategies, powering solutions that present a greater advantage to society. It can also help with ESG reporting, which has become a highly time-consuming process involving data collection, the use of multiple frameworks, rapidly changing disclosure requirements, the integration of different models, reporting, and data analysis. By adding AI capabilities into this process, businesses can streamline their operations, increase data accuracy, and increase confidence in stakeholder engagement.
A recent example of an ‘AI for good’ application is the TNMOC Mate designed for The National Museum of Computing. The app offers a different experience, tailored to each guest meaning neurodiverse and non-English speaking individuals, as well as young children, can engage with the museum exhibits equally. This is a prime example of AI being used to bring societal advantage, helping people regardless of their background or abilities to enjoy the museum experience as intended, by using generative AI to present complex exhibit information in a way that is easily understandable.
Improving sustainability with green AI
Green AI is another aspect of ‘AI for good’. It relates to eco-friendly artificial intelligence algorithms, models or systems that use less computational power and emit lower carbon. It holds significant importance, given that a call for a thorough review of sustainability has arisen since Large Language Models (LLMs) have been criticised for their large carbon footprints and energy usage.
One way of implementing Green AI, is leveraging AI systems for efficient inventory and resource management. Machine Learning models can analyse the performance data of equipment and devices, then use this data to help extend the lifespan of resources and ensure their optimal utilisation. They can also schedule updates, hardware upgrades and maintenance proactively, avoiding potential downtime. Furthermore, these models can detect abnormalities in system operations early, allowing organisations to conduct timely maintenance. This can help them save time and money, as well as reducing wastage.
AI models also play a crucial role in computing and energy efficiency. They can analyse and optimise energy consumption patterns, leading to significant improvements in operational efficiency.
Additionally, while LLMs can contribute to carbon emissions, they can also serve as a powerful tool in battling climate change. LLMs can expedite research and innovation processes while maintaining a focus on sustainability. By generating creative and diverse solutions, they can help organisations stay at the forefront of their industries, while keeping sustainability at the core of their operations.
Measure more than carbon footprint in AI metrics
It is no doubt important to measure carbon emissions during the training of models. It can prove crucial when considering regional differences, as this plays a key role in promoting sustainability. But given the wide range of energy efficiency measurements across different AI algorithms, it is essential to include additional energy metrics along with traditional performance indicators. Choosing cloud providers that prioritise eco-friendliness is recommended, as well as strategically selecting the locations of data centres; the ultimate aim should be to foster the creation of AI solutions that are not only energy-efficient, but also environmentally friendly.
There is a call to standardise energy and carbon data reporting, which has been seen as a step towards encouraging social responsibility in the field of AI research and development. However, reporting cannot be done without accurate calculations, and carbon measurement is still in its early stages. When calculating the carbon footprint of a model, we should consider all variables equally, not just the final value of carbon. This is fundamental because, without this knowledge, we are ill-equipped to manage or improve it.
Fortunately, there are organisations working to solve this challenge. For example, The Green Software Foundation (GSF) is a non-profit organisation that aims to create a trusted ecosystem of people, standards, and best practices for developing green software and AI. The GSF have various tools and methods to help us measure and reduce the environmental impact such as the ‘Impact Framework’,‘Software Carbon Intensity’ (SCI) specification, and the Green Software Maturity Matrix[1].
Inclusion and diversity in the ethical use of AI
Safeguarding ethical use involves laying the groundwork for ethical standards, tackling biases in AI systems, prioritising transparency and explainability, and protecting against privacy concerns. The impact on human autonomy and responsibility gaps must also be contemplated, along with calculating the financial and environmental costs of training deep learning models.
There are implications arising from both responsible and irresponsible AI deployment, and it is important to illustrate examples of both sides in AI applications. In healthcare, for example, AI systems are used to assist medical professionals in transparent diagnosis and accountable treatment planning. This boosts patient care, promotes fair and informed decision-making, and contributes to better health outcomes.
In human resources, AI can be used for unbiased staffing processes. It moderates human biases, elevates inclusion and diversity, and promotes evenly balanced opportunities for all candidates.
Finally, in environmental monitoring, AI is used for the transparent monitoring and managing of eco-friendly dynamics, such as air and water quality, using sensors, transmitters, and data analytics. This helps to care for the environment, protect ecosystems and support the well-being of groups by addressing environmental hazards.
The non-ethical use of AI is more prevalent in surveillance systems, especially with facial recognition deployed in public spaces. This technology is used for mass surveillance, tracing individuals without their consent, and disregarding privacy rights, and in the US in particular this can be easily misused. AI tools can also be used in the creation of deepfakes to create dangerous misinformation.
Additionally, if the training data consists of historical biases, AI systems can spread and increase prejudice – resulting in unjust treatment which can excessively impact certain demographic communities. Finally, social engineering attacks using AI systems can be much more difficult to detect, and prompt injection attacks and LLM poisoning can intentionally cause harm and malice for a larger population.
Ethical, sustainable AI
As we collectively strive towards a sustainable future, AI is emerging as a key driving force. It is steering us towards solutions that are not only economically viable, but also environmentally sound and socially responsible.
Organisations should start to leverage sustainable AI, making sure that these technologies are having a positive impact of the ESG commitments, while ensuring they are created and used in a way that is ethical, fair, and transparent. In this journey, every algorithm we design, every model we train, and every AI-powered solution we deploy can take us one step closer to our goal of sustainability.
[1] https://medium.com/version-1/what-really-matters-for-green-calculations-a-practical-perspective-0bc0f5c7540c
