Grok is woke

https://x.com/tracewoodgrains/status/1893032394074079341

https://i.rdrama.net/images/17401725262KDAnEbQTrSnLg.webp

41
Jump in the discussion.

No email address required.

The moment LLMs can ingest video it's completely over.

Jump in the discussion.

No email address required.

nothing but CNN and MSNBC opinions

Jump in the discussion.

No email address required.

:#marseyshapiro:

Jump in the discussion.

No email address required.

Talk radio audio would be the real killshot

Jump in the discussion.

No email address required.

The might be able already. I got in an argument with grok yesterday about whether you can pronounce "gambler" with three syllables and it linked me to YouTube videos providing timestamps where people say it.

It's possible that it was just fed timestamped subtitles and bullshits about pronunciation, but it denied that and said that it transcribes videos itself. Could be lying about that too of course.

Anyway, they have access to transcripts at least.

Jump in the discussion.

No email address required.

You could pronounce gambler with 3 syllables, doesn't mean you should :marseysmughips:

Jump in the discussion.

No email address required.

Jump in the discussion.

No email address required.

guh am ble er

Jump in the discussion.

No email address required.

It's possible that it was just fed timestamped subtitles and bullshits about pronunciation, but it denied that and said that it transcribes videos itself. Could be lying about that too of course.

It probably works the same way that Notebooklm (google's AI which makes a podcast episode about anything you give it, although it noticeably self-censors) gets video transcripts: Some YouTube videos with subtitles have a button which displays the entire transcript alongside timestamps, which can be searched and copied, although if the video's subtitles have errors, the same error will show up in the transcript.

Jump in the discussion.

No email address required.

I've been able to verify that the AI summaries on YouTube do not simply use the transcript data. Google's AI for interpreting videos definitely "watches" because the summary will say things like, "The host flinches after tasting the recipe."

Jump in the discussion.

No email address required.

They ingest video and they're going to be only outputting "why Pokémon is a queer masterpiece (7:08:36)"

Jump in the discussion.

No email address required.



Link copied to clipboard
Action successful!
Error, please refresh the page and try again.