stancarey,

"It's not lying, it's not telling the truth, because both of those would require some intentionality and some communicative intent, which it doesn't have."

@emilymbender talks to Michael Rosen about chatbots and the synthetic text they produce: https://www.bbc.co.uk/programmes/m001l97m

jeffreyfisher,

@stancarey @emilymbender I hadn't thought about LLMs this way before, but immediately it makes me wonder if we could say that LLM-produced text counts as "bullshit" in the specific way that Harry Frankfurt defines it, as not being concerned with truth or falsity?

stancarey,

@jeffreyfisher @emilymbender I don't think synthetic text, like output, counts as "bullshit" in the Harry Frankfurt sense. A bullshitter is indifferent to truth but knows what it is, at least hypothetically. They understand the context, even if they don't care. An LLM doesn't know what truth is, because it doesn't "know" anything.

jeffreyfisher,

@stancarey @emilymbender That's great, Stan. Maybe the programmers are the ones bullshitting. But also, do you know Plato's aporetic Hippias Minor dialogue? The question is whether Achilles, who speaks falsely without lying because he believes what he's saying (even though it's false), is worse than Odysseus, who lies but actually knows what the truth is. Maybe with LLMs we have a third option: speaking truly or falsely with no knowledge of either truth or falsity, never mind having a stake in or commitment to the truth or falsity of anything. Like the argument that "isn't even wrong," output "isn't even bullshit."

stancarey,

@jeffreyfisher Yes, just so: it's not even bullshit

grvsmth,
@grvsmth@lingo.lol avatar

@stancarey @emilymbender The lie comes from the people who present its output as truth. They have communicative intent, and they're either intending to deceive, or deceived themselves, or they just don't care whether what they're saying about the chatbots is true or not.

markvonwahlde,

@stancarey @emilymbender I'll settle for intentional misrepresentation because some human being had to put the faux human out there.

FeralRobots,
@FeralRobots@mastodon.social avatar

@stancarey @emilymbender
So, not even bullshit.

stancarey,

deleted_by_author

  • Loading...
  • FeralRobots,
    @FeralRobots@mastodon.social avatar

    @stancarey to elevate it to 'bullshit' status we'd have to conceptualize the LLM as simply an organ of the business - think of it as a system that can have something like intent. (I would argue corporations can have intent, but that's another discussion.) THEN it could be bullshit.
    ...which goes back to the critique of AI that not only does it already exist, but it has done since the first corporations were founded hundreds of years ago. (IMO it's an only slightly-facetious argument.)

    kcivey,
    @kcivey@mastodon.social avatar

    @stancarey @emilymbender I got most of the way through that sentence before I realized it wasn't about Donald Trump.

    stancarey,

    @kcivey He has a way of looming over discussion of truth in current discourse. I'm relieved to live in a country where that spectre is a little less prominent

    kcivey,
    @kcivey@mastodon.social avatar

    @stancarey Lately I've been wishing my Irish ancestors were more recent so I'd have more options if emigration becomes necessary.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • generativeAI
  • DreamBathrooms
  • mdbf
  • ethstaker
  • magazineikmin
  • GTA5RPClips
  • rosin
  • thenastyranch
  • Youngstown
  • InstantRegret
  • slotface
  • osvaldo12
  • kavyap
  • khanakhh
  • Durango
  • megavids
  • everett
  • cisconetworking
  • normalnudes
  • tester
  • ngwrru68w68
  • cubers
  • modclub
  • tacticalgear
  • provamag3
  • Leos
  • anitta
  • JUstTest
  • lostlight
  • All magazines