Remix.run Logo
Remix clone Hacker News
new | show | ask | jobsGithub
▲ kelseyfrog 3 days ago

How can you tell? They're missing the telltale sign — the em dash.

▲v4nn4 3 days ago | parent | next [-]

Good point. On a serious note, I probably overreacted, sorry about that. I have been working as a derivatives quant for a decade and thought the claim that stochastic calculus was not used/useful was ridiculous.

▲kelseyfrog 3 days ago | parent [-]

You're good bro. This is the internet; we're all here to have a good time.

▲bee_rider 3 days ago | parent | prev [-]

I hate this em dash meme. Yes, using a totally normal bit of punctuation is a sure sign that something was written by a bot.

▲kelseyfrog 3 days ago | parent [-]

How so? A human is more likely to use a hyphen, an AI an em dash. Same with quotes - a human is much more likely to use ", an AI “ and ”. Typography is a differentiating signal when it's used dis-proportionally more by one group than another.

▲bee_rider 3 days ago | parent [-]

Word processors (less of an issue for Internet comments, but worth keeping in mind) but more significantly iOS (at least) and I assume Android will just swap in an em dash where needed—it is automatic.

There is probably some signal, but be a good Bayesian; we have people saying “oh, this is a bot” when there’s a huge population of mobile users with smart keyboards that are the more likely cause.

Anyway, in general I find bot-hunting annoying. Comments should be handled as comments, if someone has made a bad argument, it should be taken down as a bad argument. If it was bot-generated, it is still there to mislead people. The advantage that bots have is that they have infinite patience and nothing better in their lives to do than argue, but there have always been people like that, so hopefully readers will be able to observe that persistence!=correctness.

▲kelseyfrog 3 days ago | parent [-]

I'm using a stock Android keyboard - it doesn't. Perhaps that's were our differing perspectives originate. I'm updating away from AI, and toward iPhone users.

EDIT: I plugged in my prior, hit rate and false alarm rates from before updating and found that my P(AI|fancy-em) = 0.09. After updating my false alarm rate, P(AI|fancy-em) now = 0.016.

▲bee_rider 3 days ago | parent [-]

Oh wow, it is a convenient feature IMO.