• 2 Posts
  • 72 Comments
Joined 1 year ago
cake
Cake day: July 3rd, 2023

help-circle






  • I think we have something contemporary to compare this sci-fi scenario with: recorded vs. live music (especially now that we can keep making exact digital duplicates as nauseam.)

    When you play a CD, it sounds the same each time (ignoring things like the equipment you’re playing through, the room, the ambient noise, etc.). Usually, the studio recording is the best, most pristine recording of a song you can get.

    But when you see the original artist performing the song live, it’s different! A good performer will make you feel like you’re experiencing something special. And the little variations, and even, imperfections, make the song even more compelling!

    It’s the CD recording of the song bad? No. It’s perfectly serviceable. It might even contain things that can’t be performed live. But it’s the same each time. And for some people, that makes it less desirable, and live performances, with all their deviations and mistakes become more desirable.

    And going back to replicated food, apart from Eddington and grandpa Sisko, I don’t remember anyone else saying replicated food was bad. Just less desirable. And even Eddington grudgingly admitted that the TV dîner he was eating in the shuttle with Sisko wasn’t that bad.








  • Is the debate being simulcast on all the major networks? I seem to remember seeing the Trump/Harris debate on ABC, CBS and NBC (just with different talking heads before and after).

    If so, ABC should broadcast the debate with fact-checking overlays (Pop-up video style?) and advertise the shit out of the fact that they’ll be doing this.


  • Ackshully…

    The TOS Enterprise, as well as the A, D and E are usually referred to as “The Enterprise”. B and C aren’t on long enough for me to remember if it applied to them as well.

    However, the NX-01 was always referred to as “Enterprise”, (without the “The”). It’s like they were making a point of it, too. So I wouldn’t count it with regards to this meme.

    So you then have to decide which ones you would count, such as the J, the one in ST:P, whatever came up on Discovery… Or whether you want to consider alternates as distinct.






  • I don’t like the idea of restricting the model’s corpus further. Rather, I think it would be good if it used a bigger corpus, but added the date of origin for each element as further context.

    Separately, I think it could be good to train another LLM to recognize biases in various content, and then use that to add further context for the main LLM when it ingests that content. I’m not sure how to avoid bias in that second LLM, though. Maybe complete lack of bias is an unattainable ideal that you can only approach without ever reaching it.