AI & Sustainability

Ronni Kahalani

�21. August 2025

Advantages and disadvantages of AI and sustainability

Let's start with some of the disadvantages of AI

The great all-embracing AI monster.

Energy and resource consumption

Data centers for training large AI models require significant amounts of energy and cooling water, which can lead to increased CO₂ emissions.

​

Solutions

• Energy-efficient AI models.
• Use renewable energy sources.

Mining & electronic waste

Computer hardware relies on rare earth elements, which can lead to environmental damage.

Short hardware lifespans contribute to electronic waste.

​

Solutions

• Reuse and recycling.
• Sustainable hardware design.
• Alternative materials.
• Sustainable mining.

Job losses due to automation

AI-powered automation in areas such as agriculture, manufacturing and logistics could replace human jobs.

Job losses could lead to economic instability and lower social sustainability.

​

Solutions

• Education / retraining.
• Innovation / adaptation.

Ethical AI and bias

AI systems can contain bias, which can lead to discrimination, unfair outcomes, and may overlook long-term ecological consequences.

Bias in data can lead to poor decisions in sustainability efforts.

​

Solutions

• Transparent datasets and algorithms.
• Ethical guidelines.

Surveillance & privacy issues

AI-based environmental monitoring could violate privacy.

Smart cities using AI could lead to mass surveillance.

​

Solutions

• Citizen involvement in decisions about AI monitoring.
• Public consultations: Cities should hold meetings and debates about the use of AI.
• Open source: AI systems used for environmental monitoring should be open for review (open-source).

Economic barriers & inequality

AI-powered sustainability solutions can be expensive for developing countries.

Unequal access to AI technology can widen global sustainability gaps.

​

Solutions

• Open-source AI and collaboration.
• Public and private investments in AI infrastructure.
• AI optimized for lower costs, which can run on smaller devices and offline, to reduce the need for expensive cloud services.

Overreliance on AI decisions

Governments and businesses can blindly trust AI recommendations without human control.

Misinterpreting AI data can lead to poor sustainability policies.

​

Solutions

• Human oversight and responsibility.
• Transparent and understandable AI.
• Independent oversight and regulation.

Greenwashing & corporate abuse

Some companies may use AI as a marketing gimmick for sustainability without any real impact.

AI can be used to maximize profits at the expense of the environment.

​

Solutions

• Stricter legislation.
• Certifications through independent organizations.
• Open data: Public access to companies' carbon footprints and energy consumption.

The benefits of AI

The future is bright.

AI and green transition in business

Companies can use AI to optimize processes and reduce waste, contributing to sustainability.

AI-powered systems can predict demand and thus reduce overproduction.

Energy Efficiency & Optimization

AI can optimize energy consumption in industries, buildings, and households, reducing waste.

Smart grids use AI to balance demand and supply of electricity efficiently.

Combating climate change

AI can improve climate models and predict environmental change more accurately.

AI-powered CO₂ capture and storage can improve emissions reduction strategies.

Sustainable agriculture & food production

AI can optimize irrigation, reduce pesticide use, and improve crop yields.

AI-powered drones and robots can reduce the need for human labor and environmental impact.

Waste management and recycling

AI-based sorting systems improve recycling rates and reduce the amount of waste in landfills.

Predictive analytics helps cities optimize waste collection.

Wildlife and biodiversity conservation

AI-based surveillance can detect poaching and illegal deforestation.

AI helps track endangered species and assess the health of ecosystems.

Sustainable transport & logistics

AI optimizes public transport and carpooling routes to reduce emissions.

Autonomous electric cars can reduce fuel consumption and traffic congestion, and can significantly increase safety.

Corporate Sustainability & ESG Compliance

AI can help companies track their carbon footprint and comply with environmental regulations.

AI-powered automation can reduce paper consumption, energy consumption, and waste.

Smart cities & urban planning

AI improves traffic management, reducing congestion and emissions.

Smart buildings use AI for energy-efficient heating, cooling, and lighting.

Water management

AI helps detect leaks and optimize water distribution.

AI-powered weather forecasts improve water conservation strategies.

AI in the circular economy

AI can help identify opportunities for reusing and recycling materials.

AI helps design products that are easier to reuse and recycle.

AI-powered marketplaces connect buyers with reused or upcycled goods.

Data-driven sustainability

By analyzing large data sets, AI can identify inefficient processes and suggest sustainable solutions.

​

Example

Application of AI in agriculture for precision farming, reducing the need for pesticides and fertilizers.

AI & UN Sustainable Development Goals

How does AI fare against the 17 UN goals on sustainability?

AI & UN Sustainable Development Goals

https://www.nature.com/articles/s41467-019-14108-y

Conclusion

AI's contribution to the world.

AI's contribution to the world

• AI is currently much more beneficial than harmful to the sustainability of the world.
• Already creating value (benefits and opportunities) for the world.
• Capable of rapid and effective adaptation to dynamic circumstances.
• Can create a much better world, but it is up to us humans.
• Negative risks can be addressed through transparency, regulation, ethical AI design, and investment in AI infrastructure and security.
• And like everything else of great value, it has a cost.

​

Questions & Feedback

Appendix

Other AI information.

World Economic Forum AI Explorer

​

​

​

​

​

​

​

​

​

​

​

​

https://intelligence.weforum.org/topics/a1Gb0000000pTDREA2?tab=publications

​

Appendiks

AI models

AI models

Grok (xAI)

• AI chatbot designed for truth, real-time information and wit.

GPT-4

• OpenAI's advanced language model excelling in text generation, reasoning, and coding.

Gemini 2.0

• Google DeepMind's multimodal AI model handling text, images, and code.

Claude 3

• Anthropic's AI focused on safe and logical text-based reasoning.

DeepSeek R1

• Open-source AI designed for enhanced reasoning and problem-solving.

Llama

• Meta's open-weight AI model optimized for text and chat applications.

AI models

Mistral

• Lightweight open-source model designed for efficiency and speed.

PaLM 2

• Google's powerful language model excelling in translation and reasoning.

Falcon 40B

• A high-performing, open-source AI model for NLP tasks.

Bard (Gemini 1)

• Google's early AI chatbot predecessor to Gemini 2.

Whisper

• OpenAI's automatic speech recognition model for transcriptions.

T5 (Text-to-Text Transfer Transformer)

• Google's NLP model treating all tasks as text generation.

AI models

BERT (Bidirectional Encoder Representations from Transformers)

• Google's model excelling in understanding context in text.

RoBERTa

• An optimized version of BERT by Facebook AI for better NLP performance.

XLNet

• Google and CMU’s model combining BERT and autoregressive approaches for better language understanding.

ALBERT

• A lighter, faster version of BERT with fewer parameters.

OPT (Open Pretrained Transformer)

• Meta’s open-source alternative to GPT.

Bloom

• Hugging Face's open-source multilingual large language model.

AI models

Stable Diffusion

• AI model by Stability AI for generating images from text.

DALL·E 3

• OpenAI’s model for creating detailed images from text prompts.

MidJourney

• A generative AI for artistic and stylized image creation.

StyleGAN

• NVIDIA’s AI model for generating high-quality synthetic images.

Runway Gen-2

• AI-powered video generation model by Runway.

CodeLlama

• Meta’s AI model for code generation and programming tasks.

AI models

Copilot (Codex)

• OpenAI’s AI assisting in code completion and suggestions.

Gemini Nano

• Google’s lightweight AI model optimized for mobile devices.

Perceiver

• DeepMind’s model handling multiple data types like images, text, and audio.

SEER (Self-Supervised)

• Meta’s AI model for training on vast amounts of unlabeled images.

WaveNet

• DeepMind’s AI model for realistic human-like speech synthesis.

Jasper

• AI-driven model optimized for automatic speech recognition.

AI models

DeepFace

• Meta’s facial recognition AI with near-human accuracy.

Sora (OpenAI)

• Text-to-video model capable of creating high-quality, realistic videos.

Make-A-Video (Meta)

• AI model generating short videos from text descriptions.

Pika Labs

• AI model generating short videos from text descriptions.

References

Referenceliste

Nature Communications: The Role of Artificial Intelligence in Achieving the Sustainable Development Goals�https://www.nature.com/articles/s41467-019-14108-y

ScienceDirect: Artificial Intelligence for Sustainability: Challenges, Opportunities, and a Research Agenda�https://www.sciencedirect.com/science/article/abs/pii/S0268401220300967

ScienceDirect: Modeling the Effects of Artificial Intelligence (AI)-Based Innovation on Sustainable Development Goals (SDGs)�https://www.sciencedirect.com/science/article/pii/S0040162523008880

ScienceDirect: Artificial Intelligence Potential for Net Zero Sustainability: Current Evidence and Prospects�https://www.sciencedirect.com/science/article/pii/S2949823624000187

Journal of Big Data: Green and Sustainable AI Research: An Integrated Thematic and Topic Modeling Analysis�https://link.springer.com/article/10.1186/s40537-024-00920-x

Environmentally Sustainable Software Design and Development: A Systematic Literature Review�https://arxiv.org/abs/2407.19901

Sustainable AI: Environmental Implications, Challenges and Opportunities�https://proceedings.mlsys.org/paper_files/paper/2022/hash/462211f67c7d858f663355eff93b745e-Abstract.html

Towards Sustainable AI: A Comprehensive Framework for Green AI�https://link.springer.com/article/10.1007/s43621-024-00641-4

Sustainability in the Software Industry: A Survey Study on the Perception, Responsibility, and Motivation of Software Practitioners�https://www.researchgate.net/profile/Dominic_Lammert/publication/379959772_Sustainability_in_the_Software_Industry_A_Survey_Study_on_the_Perception_Responsibility_and_Motivation_of_Software_Practitioners/links/6624674666ba7e2359ed1f97/Sustainability-in-the-Software-Industry-A-Survey-Study-on-the-Perception-Responsibility-and-Motivation-of-Software-Practitioners.pdf

Referenceliste

EU AI Act (2024)�https://artificialintelligenceact.eu/the-act/

EU AI Act Explorer�https://artificialintelligenceact.eu/ai-act-explorer

UN AI for Good-initiative�https://hkifoa.com/wp-content/uploads/2024/12/ai-for-good-deloitte.pdf

UN: The Sustainable Development Goals�https://www.un.org/sustainabledevelopment/sustainable-development-goals/

OECD AI Principles�https://oecd.ai/en/ai-principles

World Economic Forum: AI for Climate Action (2023)�https://www.weforum.org/stories/climate-action

Kraka og Deloitte: Mere sandsynligt, at AI gavner end skader klimaet�https://kraka.dk/wp-content/uploads/ai_er_en_hjaelp_i_den_groenne_omstilling.pdf

Carbon Footprint of ChatGpt (3. Maj 2024)�https://piktochart.com/blog/carbon-footprint-of-chatgpt