Cross-Modal Repair: Gaze and Speech Interaction for List Advancement

Generative AI Considered Harmful

“…Estimates to train GPT-3 vary from 1.29 gigawatt hours 13 to between 4 tons of CO2 (equivalent to the emissions of one non-US person in a year) to 200 tons of CO2 (equivalent to 50 non-US persons' emissions in a year)-depending on how carbon-intensive the electricity generation is regionally where the data centers are located 14 . Kasper Groes Albin Ludvigsen calculates the carbon footprint of running ChatGPT to be 23.04 kg CO2e daily, or the equivalent of roughly 76 percent of the annual carbon footprint of a Dane 15 . More research is needed to say with more certainty what the environmental impact of ChatGPT is in context of its usage, and how it compares to other information tools we use on a daily basis (e.g., search engines) to understand its real environmental cost.…”

“…As a community we have a methodological toolbox to show just how things get done in interaction, for instance through studies that draw on conversation analysis [12,21], Wizard of Oz studies [18], speech acts [16], and embodied interaction [15], to name but a few.…”