Kyle Hughes on Nostr: The carbon footprint of training Llama 3.3 was 20 passenger flights from NYC to ...
The carbon footprint of training Llama 3.3 was 20 passenger flights from NYC to London. The rest of compute happens on-demand. Why is running my Mastodon server at 60% CPU utilization a more ethical use of energy? Why is it more ethical to run pre-trained CoreML image recognition models from Apple? Why is it more ethical to use my GPU to play Call of Duty than run inference? Why aren't we picketing Apple over searchpartyd CPU utilization?
Published at
2024-12-24 15:44:53Event JSON
{
"id": "d3dc6ac1b1698da7a9910ebdacb8c9fed83d9862ec1bf0ed70a68f3b3039d49f",
"pubkey": "a41af2077c257595ebba045291d00340a8d654bf7ea68d112f1599c200f0da71",
"created_at": 1735055093,
"kind": 1,
"tags": [
[
"proxy",
"https://mister.computer/users/kyle/statuses/113708570619842565",
"activitypub"
]
],
"content": "The carbon footprint of training Llama 3.3 was 20 passenger flights from NYC to London. The rest of compute happens on-demand. Why is running my Mastodon server at 60% CPU utilization a more ethical use of energy? Why is it more ethical to run pre-trained CoreML image recognition models from Apple? Why is it more ethical to use my GPU to play Call of Duty than run inference? Why aren't we picketing Apple over searchpartyd CPU utilization?",
"sig": "5e25dbd171032557d8f6f40275272fc6fe6d5c57a4340f28ad253544cb7e82dbcae6122b9147844721fc7cf16d865936d6b2ceda3ebdeeb691d698ef859cf5f1"
}
