A high-level foundation

Here, we hope to establish a foundation for understanding GenAI systems. Instead of diving head-in, we’ll provide some context that will hopefully lead to interest and you seeking out useful critical information. We aim to foster a critical, sociotechnical perspective on AI by promoting understanding, tinkering, and critical thinking about GenAI systems. Think of this as a very beginning guide.

What is a generative AI (GenAI) system?

When you think about the word AI, what comes to mind? Is it Terminators walking through a barren wasteland, a white robot with a blue hue, or is it ChatGPT? A more formal definition states that GenAI is a computational technique capable of generating “new” content from training data. This new content could be in many forms, such as audio, video, images, text, etc. Given this definition, we can claim that ChatGPT is a generative AI system.

GenAI systems (i.e., systems using generative AI models as a main component) can be classified into five main types:

  • text generation; examples include ChatGPT, Llama app, Claude
  • image generation: text to image and vice versa; examples include DALL-E, Stable Diffusion, MidJourney
  • Audio/Music Generation: text to sound (audio, music, speech) and vice versa; examples include Suno, MusicLM
  • Code Generation: text to code; examples include Github Copilot, Claude, Vertex AI
  • speech generation: speech to speech and speech to video; examples include Sora, ResembleAI

These GenAI systems often involve the use of a large language model (LLM), which can require many resources.

AI Data Center

What impact does the creation of a GenAI system have on the environment?

Every GenAI system has come at a cost, mainly due to the GenAI model. For example, it was said that training a simple model emitted about the same amount of carbon as a trans-American flight.

These models are powered by data centers: specialized buildings equipped with computers with powerful chips needed to run the large models we use in our daily lives. These centers get extremely hot, powering all the technological activity we do, and requiring cooling systems to regulate their internal temperatures, typically using water [daxsom, 2024]. GenAI systems are particularly resource-intensive and need corresponding data centers to operate. According to the International Energy Agency, the sum of centers, over 10,000 and 30% in the United States, will contribute 6% of total electricity consumption by the end of 2026 [International Energy Agency, 2024].

AI Data Center

What we mean by Critical AI Literacy (CAIL)

Critical AI literacy is the skills and knowledge needed to develop and utilize AI technologies with a critical mindset Velander et al., 2024. CAIL caters to increasing the awareness of AI’s double-edged sword. However, CAIL helps individuals increase productivity with work processes, fosters a deeper understanding of how sociocultural structures may impact AI behavior, and can promote fair and diverse assistance to all types of communities. Some specific CAIL benefits to AI technology can include portraying diverse backgrounds, undermining traditionally marginalized voices in designs, highlighting the environment, and understanding repercussions from over-dependence on GenAI systems [Veldhuis, 2024].

The introduction of ChatGPT has brought salient changes to the educational landscape for students and teachers alike. Teachers have begun using these systems to create schedule plans and rubrics while having to deal with increased plagiarism [Huang, 2024]. Those with a lackluster level of CAIL will not be able to pinpoint that chatbots conflate information with absolute truth and disseminate harmful information targeted at historically marginalized groups Tanksley, 2024.

An opportunity to tinker with some code

By now, you should have a basic understanding of

  • What is a GenAI system? What is the main component, and what are some related costs from the creation of a system
  • What is CAIL, how is it beneficial, and why it important in this age of rapid AI development

If you would like to see some example code and potentially tinker with a system, then follow this link.

If you would like to continue to learn more about environmental impacts (and get to the chatbot you’ll be able to use for the design competition), follow this link. (or just click on MEPO 2026 -> AI & The Environment on the left hand menu)