Record🤔

uri:

"at://did:plc:x2nsupeeo52oznrmplwapppl/app.bsky.feed.post/3l4gsfkcidl2r"

cid:

"bafyreifboy7vkdul6tn7xigorpqkoepf7zwqww4qcfeptdu6bals3y5pby"

value:

text:

"I loathe to resume the LLM discourse, but today's @washingtonpost.com article is a great example of why I struggle to take the energy cost discussion seriously.

Article: one inference is 519ml, 0.14kWh

Linked paper: one inference is 17ml, 0.004kWh

🤷‍♂️

www.washingtonpost.com/technology/2..."

$type:

"app.bsky.feed.post"

embed:

$type:

"app.bsky.embed.external"

external:

uri:

"https://www.washingtonpost.com/technology/2024/09/18/energy-ai-use-electricity-water-data-centers/"

thumb:

View blob content

$type:

"blob"

ref:

$link:

"bafkreibvkyxhj6b6tlvraipuc6qjgvokeeuadtwoei3ncdnqkht3pdpj4q"

mimeType:

"image/jpeg"

size:

918649

title:

"A bottle of water per email: the hidden environmental costs of using AI chatbots"

description:

"AI needs a lot of electricity and water to stay cool in data centers. We break down the toll prompt-by-prompt to show the scale of AI’s environmental needs."

langs:

• "en"

facets:

• $type:

  "app.bsky.richtext.facet"

  index:

  byteEnd:

  69

  byteStart:

  50

  features:

  • did:

    "did:plc:k5nskatzhyxersjilvtnz4lh"

    $type:

    "app.bsky.richtext.facet#mention"

• index:

  byteEnd:

  304

  byteStart:

  266

  features:

  • uri:

    "https://www.washingtonpost.com/technology/2024/09/18/energy-ai-use-electricity-water-data-centers/"

    $type:

    "app.bsky.richtext.facet#link"

createdAt:

"2024-09-18T15:02:08.167Z"