AI, DATA and ROBOTICS FOR SUSTAINABILITY

The double-edged sword

The European Convergence Summit 2024 focused on the Impact of AI, Big Data and Robotics on CO2 reduction. Our President Dr Monique Calisti gave a keynote presentation opening the ECS 2024 Panel dedicated to exploring positive and negative impacts of AI, Data and Robotics (ADR) technology on the environment. An overview of some main takeaways is provided below and the slides can be downloaded from HERE.

MAIN ENVIRONMENTAL CHALLENGES HUMANITY MUST FACE

According to the World Economic Forum’s Global Risks Report 2024, the major and most urgent environmental challenges humanity must address are:

Pollution: air, water, and soil contamination.
Biodiversity loss: habitat destruction, species extinction.
Resource depletion: overconsumption of natural resources.
Climate change: rising global temperatures, extreme weather events.
Waste management: increasing waste volumes and inadequate disposal methods.

These environmental problems have far-reaching consequences, not only for the natural world but also for human societies and economies.

THE GREEN PROMISE OF ICT AND IN PARTICULAR OF ADR TECHNOLOGIES

ICT and in particular ADR technologies, either isolated or combined, have a huge potential to help address several environmental challenges across various sectors, by contributing to the improvement of several impactful activities, such as:

Optimising Processes and Supply Chains
Improving Environmental Monitoring and Analysis
Optimising Resource Management and Efficiency
Supporting Better Informed and Timely Decision Making

ADR technologies and solutions are already in action for the planet in many ways, by allowing for instance predictive analytics for climate modelling and weather forecasting; optimization of energy usage in buildings and industrial processes; advanced CO2/GHG monitoring; informed decision-making for resource management and increased efficiency.

In particular, ADR solutions have the potential to cut global Greenhouse Gas (GHG) emissions in other sectors by 15% by 2030: AI alone, for instance, is capable of shaving 5 to 10% off of global GHG emissions.

REBOUND EFFECT AND ECOLOGICAL COSTS

However, although the positive impact of ADR solutions across a variety of industries is huge, their negative impact cannot be neglected when aiming at Net Zero.

High Energy Consumption and Carbon Emissions:

Training complex AI models, especially large ones, requires a massive amount of electricity, which today is still produced mostly from fossil fuels. With adoption of machine learning and generative AI at scale, the inference use of already trained models now itself requires extreme computational power and associated energy.
Data centres are among the largest consumers of electricity globally. If data centres were a country, they would rank as the fifth-largest energy consumer. If this energy is not green, clearly there is a huge consequent amount of CO2 emissions.
Robots, particularly those in industrial settings, consume energy during operation. If this energy comes from non-renewable sources, it contributes to CO2 emissions.

Global resource consumption:

The manufacturing of specialised hardware like GPUs needed for AI training and inference can have a significant footprint due to resource extraction and energy use.
Data centres require substantial physical resources for construction and operation, including water and rare earth metals, which have a substantial extraction and processing impact.
The production of robots requires a significant amount of resources, including metals and plastics, which can lead to environmental degradation and resource depletion.

Negative impact on Natural Ecosystems:

The resources required by AI systems can lead to habitat destruction and biodiversity loss as natural areas are exploited for materials.
Cooling systems for data centres consume vast amounts of water, which can lead to water scarcity and affect local ecosystems, but also induce air pollution.
The use and disposal of robots can create pollution. For example, robots used in mining or bomb disposal may release harmful substances into the environment.
Robotics can disturb natural landscapes and ecosystems, whether through their deployment in natural settings or due to the infrastructure needed to support them.
Some robots can produce radiation that can disturb wildlife and humans, and electromagnetic radiation from robots can potentially harm plants and animals.

E-Waste:

With rapid technological advancements, hardware becomes obsolete quickly, leading to increased electronic waste that can be harmful if not properly recycled.
The Global E-waste Monitor 2023 by the UN University reported a record high of 57.4 million metric tons of e-waste generated globally in 2021. Only 17.4% of this was documented as properly collected and recycled.

DECARBONISING ADR FOR NET ZERO - WHAT DOES IT TAKE TO GET THERE

To make sure that the positive effects of digital technologies outweigh their negative impact, it is necessary to act at several levels:

Technological advancements – sustainability by design.
Policy and Regulation – to standardise, prevent and protect while providing incentives and sharing best practices.
Individual and Organisational behaviour – educate and promote responsible choices.
Collaboration and Awareness – multi-stakeholder and global approach, raising understanding at individual and organisational level.

As presented in the Keynote, when zooming into the decarbonisation agenda for ADR technologies, there are several areas that need further work and a multi-stakeholders and multidisciplinary collaborative approach, including energy-efficient algorithms; responsible and optimised data collection and management; greener networks, data centres, edge nodes and devices; life-cycle assessment of ADR systems; biodegradable and sustainable materials for robotics engineering, etc.

DRIVING APPROPRIATE CHOICES - AN HOLISTIC AND RADICAL APPROACH

In particular, to drive appropriate choices on “if and how to digitalise for the planet”, we need:

Credible end-to-end quantitative assessments of the net environmental impact with and without a specific technological solution. This requires further investigation and work especially in terms of defining the appropriate metrics and what socio-technical consequences derive from choosing one metric over another.
A conceptual framework that covers all sustainability dimensions, spanning across economic, societal and environmental aspects. This means we can reframe the problem as a multi-objective Pareto optimisation one: different “optimal” solutions can be found by jointly considering several aspects related to sustainability, while embracing a complexity mindset that accounts, for example, for concepts such as situatedness, path-dependency, or finite budgets for non-renewable resources (or depletion rate limits for renewable ones).

Notice that unfortunately, the digital transformation carries an intrinsic risk of severing our understanding and decision-making processes from the unique and situated reality. ADR technologies (especially AI) can exacerbate this risk due to the numerical nature of ML, in combination with the strong forces and interests towards the financialisation and fungibility of natural resources and ecosystems.

This is why the intent to decarbonise ADR technologies results in some practical guidelines that should be applied to ADR system development and operation, such as, for example: 1) couple data extraction and processing with narratives, traces, and explainability; 2) preserve the identity of real-world counterparts of digital entities; 3) keep each individual goal or optimization criterion separate and human-recognisable, avoiding scalarisation pitfalls; and 4) couple numerical prediction or recommendations with alternative sanity checks and refutation procedures – some AI methods such as causal inference can offer that.

SUSTAINABILITY IS NOT ENOUGH - LET’S MOVE TO REGENERATION

Sustainability focuses on minimising our negative impact on the environment, essentially taking what we need without depleting resources. Regeneration, however, goes a step further: it aims to restore and improve the environment, leaving it in a better state than we found it. This is only possible by enforcing a change of perspective in several ways:

Focus: Sustainability emphasises maintaining a balance, while regeneration emphasises improvement. Imagine a forest: sustainability might involve harvesting trees responsibly to not cut down more than can grow back. Regeneration would involve planting new trees and improving the forest’s health for future generations.
Impact: Sustainability is about reducing our footprint. Regeneration is about creating a positive footprint, actively enhancing the environment, leveraging the dynamics and resilience that are embedded in all natural systems: not fixing/improving the environment as a passive target of human actions, but operating with and within suitable natural systems for their most desirable evolution and balance.
Long-term view: Sustainability is crucial for the future, but regeneration takes a more proactive approach. It acknowledges that simply maintaining the status quo will not solve environmental problems.

In this respect, our vision is that the powerful trio of decarbonised AI, Big Data and Robotics can become the enablement architects of a regenerative future, analysing vast datasets to optimise resource use, guide restoration efforts and allow humans and robots to heal and improve our planet.

AI, DATA and ROBOTICS FOR SUSTAINABILITY

The double-edged sword

MAIN ENVIRONMENTAL CHALLENGES HUMANITY MUST FACE

THE GREEN PROMISE OF ICT AND IN PARTICULAR OF ADR TECHNOLOGIES

REBOUND EFFECT AND ECOLOGICAL COSTS

DECARBONISING ADR FOR NET ZERO - WHAT DOES IT TAKE TO GET THERE

DRIVING APPROPRIATE CHOICES - AN HOLISTIC AND RADICAL APPROACH

SUSTAINABILITY IS NOT ENOUGH - LET’S MOVE TO REGENERATION

Share with Your Network

Privacy settings

With the slider, you can enable or disable different types of cookies:

This website will:

This website won't:

This website will:

This website won't:

This website will:

This website won't:

This website will:

This website won't:

This website will:

This website won't: