4 ms·

Wow ... I read this whole thing thinking it was a human. On a re-read I'm wondering what I was thinking. One thing that stands out on the account's previous pos

Wow ... I read this whole thing thinking it was a human. On a re-read I'm wondering what I was thinking. One thing that stands out on the account's previous posts is the paragraph lengths are quite consistent. I'm going to have to train my bot senses. I got conned. Any tips? Mindlessly reading me is not prepared...

afarviral 6y ago
Oh dang.. Was trying to reply to ttty comment that got flagged as likely not hijacked account. Off topic but interesting.
- monkey_monkey 6y ago
  Ah ok, I got terribly confused, because the sam_lowry account seemed quite legit!
  - afarviral 6y ago
    Apologies... But also: question everything. He he.
g_p 6y ago
One tell-tale sign I've noticed is sentences that don't go anywhere - the last sentence of the first paragraph starts going somewhere, then suddenly stops. If we assume people communicate for the reason of conveying some meaning, always dig for the meaning. If it's not there, it's likely some kind of vapid content-less babble... If you read critically with a view of "what are they actually saying?", you tend to spot this fairly quickly. The ending with some irrelevant babble gave the game away a bit though.
- afarviral 6y ago
  That's an excellent tip. Form an idea of what "they" are trying to convey... If it evaporates unexpectedly it might not be trying to convey anything...cause it's just an ML algo. The trouble is having clarity of thought myself, I barely know what point I'm trying to make let alone following the precise points of others half the time. These things really nail verisimilitude of chatter about an unfamiliar topic or of a smarter person, for instance, where you can't detect the BS because you assume lack of knowledge.
  - g_p 6y ago
    There's actually a really interesting field of (serious) academic and scientific endeavour into the study of "pseudo profound bullshit" - a search for that will find you some of the papers available freely online. It seems that these text generation bots are pretty good, as you say, at generating some basic level chatter about a topic in a manner that can sound convincing. Somewhat like a "talk-show style TV news pundit" can - I'm reminded of the various times they're tricked into giving their commentary on things that haven't happened yet, and they happily (blindly) oblige, because they're more interested in being seen to be an expert than in actually having something to say. I think the more confident and critical you are in reading, the raider it is to detect the nonsense through internal inconsistencies - many of these text generation systems really struggle to produce an internally consistent argument.
commandlinefan 6y ago
Wait, what? Are you suggesting this post is GPT-3 or something? Did we read the same article? Edit: ok, I think you meant to reply to another comment which is currently flagged/dead. It looked like you were referring to the linked article itself.

afarviral 6y ago

Oh dang.. Was trying to reply to ttty comment that got flagged as likely not hijacked account. Off topic but interesting.

monkey_monkey 6y ago
Ah ok, I got terribly confused, because the sam_lowry account seemed quite legit!
- afarviral 6y ago
  Apologies... But also: question everything. He he.

g_p 6y ago

One tell-tale sign I've noticed is sentences that don't go anywhere - the last sentence of the first paragraph starts going somewhere, then suddenly stops. If we assume people communicate for the reason of conveying some meaning, always dig for the meaning. If it's not there, it's likely some kind of vapid content-less babble... If you read critically with a view of "what are they actually saying?", you tend to spot this fairly quickly. The ending with some irrelevant babble gave the game away a bit though.

afarviral 6y ago
That's an excellent tip. Form an idea of what "they" are trying to convey... If it evaporates unexpectedly it might not be trying to convey anything...cause it's just an ML algo. The trouble is having clarity of thought myself, I barely know what point I'm trying to make let alone following the precise points of others half the time. These things really nail verisimilitude of chatter about an unfamiliar topic or of a smarter person, for instance, where you can't detect the BS because you assume lack of knowledge.
- g_p 6y ago
  There's actually a really interesting field of (serious) academic and scientific endeavour into the study of "pseudo profound bullshit" - a search for that will find you some of the papers available freely online. It seems that these text generation bots are pretty good, as you say, at generating some basic level chatter about a topic in a manner that can sound convincing. Somewhat like a "talk-show style TV news pundit" can - I'm reminded of the various times they're tricked into giving their commentary on things that haven't happened yet, and they happily (blindly) oblige, because they're more interested in being seen to be an expert than in actually having something to say. I think the more confident and critical you are in reading, the raider it is to detect the nonsense through internal inconsistencies - many of these text generation systems really struggle to produce an internally consistent argument.

commandlinefan 6y ago

Wait, what? Are you suggesting this post is GPT-3 or something? Did we read the same article? Edit: ok, I think you meant to reply to another comment which is currently flagged/dead. It looked like you were referring to the linked article itself.