Record🤔

uri:

"at://did:plc:oulaoye36ffniyupeeljrm6k/app.bsky.feed.post/3l46kgzoygy2y"

cid:

"bafyreidk2sirkkfmotgr3zfphpv3o3keabfvz7rhl2phe6eqgq7ahkaqgq"

value:

text:

"Wonder if the alignment problem described is just a sales pitch or a real problem?

www.transformernews.ai/p/openai-o1-..."

$type:

"app.bsky.feed.post"

embed:

$type:

"app.bsky.embed.external"

external:

uri:

"https://www.transformernews.ai/p/openai-o1-alignment-faking?utm_campaign=post&utm_medium=web"

thumb:

View blob content

$type:

"blob"

ref:

$link:

bafkreih33zqvlo6zlujwvrnngqhwc4hkmczeprmmquch3s5jihpfl5wc4u

mimeType:

"image/jpeg"

size:

515529

title:

"OpenAI's new models "instrumentally faked alignment""

description:

"The o1 safety card reveals a range of concerning capabilities, including scheming, reward hacking, and biological weapon creation."

langs:

• "sv"

facets:

• index:

  byteEnd:

  122

  byteStart:

  84

  features:

  • uri:

    "https://www.transformernews.ai/p/openai-o1-alignment-faking?utm_campaign=post&utm_medium=web"

    $type:

    "app.bsky.richtext.facet#link"

createdAt:

"2024-09-15T08:18:30.198Z"