Issues¶

Environmental Issues¶

Sources of carbon footprint

• train models
• data collection
• inference
• creating new chips

Types¶

• Embodied carbon
• Operational carbon

Not just the energy cost of running computers, but also energy required to build them

Solutions¶

• Efficiency of models
• Software infra to promote development of efficiency
• AutoML instead of grid search
• Easy-to-use APIs for efficient models
• Efficient computer chips
• Efficient datacenters
• Microsoft: underwater servers
  • 8 x more reliable, using Nitrogen, etc
  • No need for fresh water for cooling
• Use renewable energy
• Transitioning to 100% renewables will not eliminate the carbon footprint of chips
• 
• Stop planned obscelence

4 M’s of AI Efficiency¶

1. Model
2. Machine
3. Mechanism
4. Map: Location

Jevon’s Paradox/Effect¶

Improved efficiency in resource utilization increases the total consumption of resources, due to increased rate of consumption from increased demand

Hence, definition of “efficiency” is important

PUE¶

Power Usage Effectiveness

Lower is better

Typically \(\in [1.1, 1.6]\)

How the power input is being used for compute & for other supporting consumption such as cooling, lighting, etc. $$ \begin{aligned} \text{PUE} &= \dfrac{\text{Total Energy}}{\text{Productive Energy}} \ &= 1 + \dfrac{\text{Overhead}}{\text{Productive Energy}} \end{aligned} $$

Ethical Issues¶

New technologies are disruptive, not neutral

AI can - change current practices - influence human decisions - regulate human behavior

Responsible AI¶

• Researchers and engineers should
  • be proactive - not reactive - about ethics
  • anticipate how their work will impact society
  • incorporate human values throughout all stages of product's lifecycle
• Increased public trust means
  • Improved marketability
  • Higher product adoption

Abstraction¶

Researchers/Engineers often abstracted away from application they’re working on, hence not aware of bad things that can be used for

Data Ownership¶

Data governance

Trade-off: Privacy vs Progress

Model monetization trained on consumer data, eg: GitHub Copilot

Data Bias¶

Dataset curation

Federated learning

Research Inequality¶

• Inequitable access to computing resources
• Inequitable access to datasets

Bias¶

Solution

Comments