Experimental browser for the Atmosphere
Let's consider... A larger model like GPT-4 would burn around 0.18Wh per 1,000 tokens, i.e. about four pages. It would take a human 5 minute to read that much content. Let's say 3 minutes, and we consider a cost for doing so 10% of the total energy demand for a person for that time. Reasonable...
May 8, 2025, 11:25 AM
{ "uri": "at://did:plc:djznhtuxa6mqmne7s3bpgj3i/app.bsky.feed.post/3lonsi2qjo22a", "cid": "bafyreiffjev5qntu4z3uwl7bnw7ft3y6i5loctpya5zmfmhckkt7zadrpi", "value": { "text": "Let's consider... A larger model like GPT-4 would burn around 0.18Wh per 1,000 tokens, i.e. about four pages. \n\nIt would take a human 5 minute to read that much content. Let's say 3 minutes, and we consider a cost for doing so 10% of the total energy demand for a person for that time. Reasonable...", "$type": "app.bsky.feed.post", "embed": { "$type": "app.bsky.embed.external", "external": { "uri": "https://anuragsridharan.substack.com/p/we-can-use-tokens-to-track-ais-carbon", "thumb": { "$type": "blob", "ref": { "$link": "bafkreibbltitrj53sreloefdpbvbowmxicvoer65nfrlfag4trhhvinwiq" }, "mimeType": "image/jpeg", "size": 354188 }, "title": "We can use \"tokens\" to track AI's carbon emissions: here's how", "description": "If we make some one time calculations, I think we can use something people already track to extrapolate carbon emissions easily" } }, "langs": [ "en" ], "reply": { "root": { "cid": "bafyreiarldq3r6egyedjuelr7oryu5zrednytfu4wrhb33a2hxxs3mcqyi", "uri": "at://did:plc:djznhtuxa6mqmne7s3bpgj3i/app.bsky.feed.post/3lonp6lnhz224" }, "parent": { "cid": "bafyreicduttqlapte524fspfz5bjcu5uvmmnxms4sf2i7rtqw525oft354", "uri": "at://did:plc:djznhtuxa6mqmne7s3bpgj3i/app.bsky.feed.post/3lonsajcp422d" } }, "createdAt": "2025-05-08T11:25:40.240Z" } }