Record🤔

uri:

"at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3kl3gwyc6ze27"

cid:

"bafyreiauf7xiw36h2lpd3t7rdea25rjrqveprajfisxklhixwg2qcg26yq"

value:

tags:

text:

"Genuinely important paper. Right now, generative models are trained in a way that makes them basically incapable of representing human *differences*. The authors explore different ways of changing that, and acknowledge tensions between them. #MLSky 🤖 arxiv.org/abs/2402.05070"

$type:

"app.bsky.feed.post"

embed:

$type:

"app.bsky.embed.external"

external:

uri:

"https://arxiv.org/abs/2402.05070"

thumb:

View blob content

$type:

"blob"

ref:

$link:

bafkreiaaivijmmfc5uof6cdxta6i72kdy73r4hsafwcklomig4bf3qxcza

mimeType:

"image/png"

size:

35487

title:

"A Roadmap to Pluralistic Alignment"

description:

"With increased power and prevalence of AI systems, it is ever more critical that AI systems are designed to serve all, i.e., people with diverse values and perspectives. However, aligning models..."

langs:

• "en"

facets:

• $type:

  "app.bsky.richtext.facet"

  index:

  byteEnd:

  248

  byteStart:

  242

  features:

  • tag:

    "MLSky"

    $type:

    "app.bsky.richtext.facet#tag"

• $type:

  "app.bsky.richtext.facet"

  index:

  byteEnd:

  278

  byteStart:

  254

  features:

  • uri:

    "https://arxiv.org/abs/2402.05070"

    $type:

    "app.bsky.richtext.facet#link"

createdAt:

"2024-02-10T18:29:33.715Z"