▲ 121 ▼ JDPON Don has caught on that commie memes require lots of text (thelemmy.club) submitted 4 days ago by corgiwithalaptop@hexbear.net [M] to c/chapotraphouse@hexbear.net 55 comments fedilink hide all child comments
[–] EmmaGoldman@hexbear.net 25 points 4 days ago (8 children) Em-dash and verbosity, name a more AI duo. permalink fedilink source hideshow 8 child comments replies: [–] g_mer@hexbear.net 10 points 4 days ago (5 children) It’s not that em-dashes necessarily indicate AI — it’s that it’s written in an AI voice. permalink fedilink source parent hideshow 5 child comments replies: [–] CloutAtlas@hexbear.net 15 points 4 days ago (4 children) I have people questioning if my posts elsewhere are AI (even shit from like 2017), because I sometimes use "It's not x, it's y" as a format but that's because my first language is Chinese and "不是x,而是y" is a perfectly normal way of phrasing things. permalink fedilink source parent hideshow 4 child comments replies: [–] invalidusernamelol@hexbear.net 10 points 4 days ago* (2 children) AI voice is actually just the voice of the millions of non-native English speakers that were mined for human reenforcement during training. For me to peg it as AI, there needs to be a reason for existence (ad, propaganda, trolling, etc.), and at least 3 of the markers. Em-dash, rule of 3, "not x, but y", excessive emoji, or a clean conclusion sentence/paragraph. Also no edits! No one ever edits an LLM post. permalink fedilink source parent hideshow 2 child comments replies: [–] fox@hexbear.net 7 points 3 days ago There's another marker which is vaguer but really hard to miss once you get the sense for it: punchiness. Can't really give any clear examples but it's when it'll do a series of terse sentences, using "strong" words to make them land harder. "One crack, many impacts" when describing how thunder rolls. permalink fedilink source parent [–] Le_Wokisme@hexbear.net 6 points 3 days ago the way chatboxes use emoji is something i've never seen actual people do permalink fedilink source parent [–] stalin420@hexbear.net 6 points 3 days ago Probably the AI usage is due to so many “Zhang, et al” science papers being in the training data permalink fedilink source parent [–] christian@hexbear.net 7 points 4 days ago It's SI now, dum-dum. permalink fedilink source parent [–] aanes_appreciator@hexbear.net 6 points 4 days ago not enough bullet points, markdown tables, emojis, and unnecessarily short utterances permalink fedilink source parent
[–] g_mer@hexbear.net 10 points 4 days ago (5 children) It’s not that em-dashes necessarily indicate AI — it’s that it’s written in an AI voice. permalink fedilink source parent hideshow 5 child comments replies: [–] CloutAtlas@hexbear.net 15 points 4 days ago (4 children) I have people questioning if my posts elsewhere are AI (even shit from like 2017), because I sometimes use "It's not x, it's y" as a format but that's because my first language is Chinese and "不是x,而是y" is a perfectly normal way of phrasing things. permalink fedilink source parent hideshow 4 child comments replies: [–] invalidusernamelol@hexbear.net 10 points 4 days ago* (2 children) AI voice is actually just the voice of the millions of non-native English speakers that were mined for human reenforcement during training. For me to peg it as AI, there needs to be a reason for existence (ad, propaganda, trolling, etc.), and at least 3 of the markers. Em-dash, rule of 3, "not x, but y", excessive emoji, or a clean conclusion sentence/paragraph. Also no edits! No one ever edits an LLM post. permalink fedilink source parent hideshow 2 child comments replies: [–] fox@hexbear.net 7 points 3 days ago There's another marker which is vaguer but really hard to miss once you get the sense for it: punchiness. Can't really give any clear examples but it's when it'll do a series of terse sentences, using "strong" words to make them land harder. "One crack, many impacts" when describing how thunder rolls. permalink fedilink source parent [–] Le_Wokisme@hexbear.net 6 points 3 days ago the way chatboxes use emoji is something i've never seen actual people do permalink fedilink source parent [–] stalin420@hexbear.net 6 points 3 days ago Probably the AI usage is due to so many “Zhang, et al” science papers being in the training data permalink fedilink source parent
[–] CloutAtlas@hexbear.net 15 points 4 days ago (4 children) I have people questioning if my posts elsewhere are AI (even shit from like 2017), because I sometimes use "It's not x, it's y" as a format but that's because my first language is Chinese and "不是x,而是y" is a perfectly normal way of phrasing things. permalink fedilink source parent hideshow 4 child comments replies: [–] invalidusernamelol@hexbear.net 10 points 4 days ago* (2 children) AI voice is actually just the voice of the millions of non-native English speakers that were mined for human reenforcement during training. For me to peg it as AI, there needs to be a reason for existence (ad, propaganda, trolling, etc.), and at least 3 of the markers. Em-dash, rule of 3, "not x, but y", excessive emoji, or a clean conclusion sentence/paragraph. Also no edits! No one ever edits an LLM post. permalink fedilink source parent hideshow 2 child comments replies: [–] fox@hexbear.net 7 points 3 days ago There's another marker which is vaguer but really hard to miss once you get the sense for it: punchiness. Can't really give any clear examples but it's when it'll do a series of terse sentences, using "strong" words to make them land harder. "One crack, many impacts" when describing how thunder rolls. permalink fedilink source parent [–] Le_Wokisme@hexbear.net 6 points 3 days ago the way chatboxes use emoji is something i've never seen actual people do permalink fedilink source parent [–] stalin420@hexbear.net 6 points 3 days ago Probably the AI usage is due to so many “Zhang, et al” science papers being in the training data permalink fedilink source parent
[–] invalidusernamelol@hexbear.net 10 points 4 days ago* (2 children) AI voice is actually just the voice of the millions of non-native English speakers that were mined for human reenforcement during training. For me to peg it as AI, there needs to be a reason for existence (ad, propaganda, trolling, etc.), and at least 3 of the markers. Em-dash, rule of 3, "not x, but y", excessive emoji, or a clean conclusion sentence/paragraph. Also no edits! No one ever edits an LLM post. permalink fedilink source parent hideshow 2 child comments replies: [–] fox@hexbear.net 7 points 3 days ago There's another marker which is vaguer but really hard to miss once you get the sense for it: punchiness. Can't really give any clear examples but it's when it'll do a series of terse sentences, using "strong" words to make them land harder. "One crack, many impacts" when describing how thunder rolls. permalink fedilink source parent [–] Le_Wokisme@hexbear.net 6 points 3 days ago the way chatboxes use emoji is something i've never seen actual people do permalink fedilink source parent
[–] fox@hexbear.net 7 points 3 days ago There's another marker which is vaguer but really hard to miss once you get the sense for it: punchiness. Can't really give any clear examples but it's when it'll do a series of terse sentences, using "strong" words to make them land harder. "One crack, many impacts" when describing how thunder rolls. permalink fedilink source parent
[–] Le_Wokisme@hexbear.net 6 points 3 days ago the way chatboxes use emoji is something i've never seen actual people do permalink fedilink source parent
[–] stalin420@hexbear.net 6 points 3 days ago Probably the AI usage is due to so many “Zhang, et al” science papers being in the training data permalink fedilink source parent
[–] aanes_appreciator@hexbear.net 6 points 4 days ago not enough bullet points, markdown tables, emojis, and unnecessarily short utterances permalink fedilink source parent