Notes: Environmental Impacts

Contents: Environmental impacts generally, environmental impacts of digital technologies, case study on NLP, LLM, AI, other ethical and social considerations, positive impacts, and what those working with digital technologies can do to reduce negative impacts.
Motivation: It's important to think about environmental impacts.
Global Temperature Change: The average global temperature has been rising.

Solar radiation passes through the atmosphere.
Some is reflected by the atmosphere and Earth's surface.
Solar energy is absorbed by the Earth's surface, warming it and causing the emission of longwave (infrared) radiation back to the atmosphere.
Greenhouse gases absorb and re-emit some infrared radiation, warming the Earth's surface and troposphere.
Some infrared radiation passes through the atmosphere and is lost in space.

Data Centers: Use significant amounts of energy for processing and cooling.
Energy Consumption: The energy consumption of ICT (Information and Communication Technology) is a concern, with figures varying widely.
- Estimates for ICT's share of global electricity range from 1.8-3.9% in 2017 to 8-10% by 2030 (excluding manufacturing).
- This is compared to other sectors like lighting, aviation, and shipping.
Lifecycle of Digital Technologies: Impacts extend from manufacturing and transport to use and disposal.
- Manufacturing: Resource extraction, energy-intensive processes, water consumption, hazardous waste, and social costs (e.g., child labor, unsafe conditions).
- Transport: Fuel use and emissions.
- Use: Electricity consumption by devices and data centers.
- Disposal: E-waste (toxic materials, low recycling rates, informal recycling leading to pollution).
Key Insight: The environmental impacts of digital technologies are significant across their entire lifecycle.

Carbon Footprint of AI: Training large AI models can have a substantial carbon footprint due to high computational demands.
- Training a large Transformer model (like those used for NLP) can emit as much carbon as five cars over their lifetime, including the energy from "green" data centers.
Comparison: Energy consumption for training AI models can be much higher than for human activities like cross-Atlantic flights.

Definition: Increasing efficiency in resource use can lead to an overall increase in resource consumption.
Explanation: Technological advancements that make a resource more efficient or cheaper to use can lead to increased demand for that resource, thereby offsetting or even increasing total consumption.
Digital Context: Digital technologies designed to improve efficiency can paradoxically increase energy demand by enabling more computation, data storage, and device usage.
- Example: Cloud computing's efficiency allows more people to use it, increasing overall energy consumption.
Implications: Efficiency alone is not enough; demand reduction and a shift in economic models are also needed.

Misinformation: The spread of climate change denial or misleading information online.
Digital Divide: Inequalities in access to technology can hinder climate action or adaptation efforts for vulnerable communities.
Surveillance: Use of digital tools for environmental monitoring can raise privacy concerns.

Monitoring & Data Collection: Sensors and IoT devices for environmental monitoring (e.g., air quality, deforestation, water levels).
Efficiency & Optimization: Smart grids, precision agriculture, and optimized logistics can reduce resource use.
Awareness & Education: Online platforms for raising awareness, sharing information, and mobilizing climate action.
Research & Modeling: AI and supercomputing for climate modeling, predicting environmental changes, and developing solutions.
Renewable Energy Management: Digital tools for managing and optimizing renewable energy sources.

Reduce Demand: Focus on reducing the demand for energy and computing technologies.
- This means understanding and mitigating rebound effects.
Sustainable Design: Prioritize energy efficiency in hardware and software design.
Ethical AI Development: Consider environmental impact in AI training and deployment.
Circular Economy: Promote repair, reuse, and recycling of e-waste.
Advocacy & Policy: Support policies that regulate environmental impact and promote sustainability in the tech sector.
Focus on Thriving, Not Just Growth: Challenge the assumption that economic growth is always beneficial and prioritize societal and ecological well-being.
- Businesses need to internalize environmental costs.
Innovation: Innovate responsibly to reduce negative impacts.