Ethics and AI - Energy Consumption

Question:

What the is energy consumption footprint of using generative AI compared to conventional digital platforms?

Inquiry Length:

Mini, 6 weeks

Why this Question is Important

There is much speculation around how much energy is being used to train and use Generative AI models. As part of the process of assessing whether AI is a viable resource for artists and designers, energy consumption is a major factor - particularly for individuals or organizations that orient their mission around environmental sustainability. Beyond the question of personal ethics, energy consumption is also likely to be a significant factor in how usage of these models will be priced.

Criteria for Successful Submission

This submission should be in the form of a written article containing the following:

• A detailed analysis of how much energy is consumed in the following components of using generative AI
  • Training the models
  • Server-side energy consumption of inference processes
  • Client-side energy consumption of inference processes
• Evidence of an attempt to produce two artifacts of similar quality.
  • One using primarily generative AI based tools from ideation to final production
  • One using primarily conventional digital tools from ideation to final production
• Comparison of both methods
  • Energy consumed for the artifacts produced as part of this inquiry
  • An extrapolation of how much energy would be consumed by MIAD by both methods
• Evidence of the research being conducted in such a way that insulates the analysis from personal bias.