GirlChat #728239

Start A New Topic!  Submit SRF  Thread Index  Date Index  


Posted by jd420 on Thursday, February 07 2019 at 10:49:57AM
In reply to You have amused me tremendously, my friend! posted by Gimwinkle on Thursday, February 07 2019 at 08:53:06AM

Now, for your case study: prove that this response to your post was either by a human or by a machine.


Did ya miss that part earlier? Nothing can attain 'proof' due to the black swan problem; only experimental falsification is possible.

Grammar's a bit better, though.

Of course, there probably are many spots where grammar failed a bit because I didn't read it all for accuracy.

Yes. Yes there were. ;)

It was one of the two giveaways - "that doesn't even english," due to grammar, and the other being "it just contradicted itself directly."

You have got to be one helluva smart guy to catch those pieces that apply.

175-195 swing on the testing. It's not worth a damn; I haven't fixed the world yet.

Normally, people are "supposed" to think, "what the hell is all that?" and give up.

I considered it, briefly. And the constant juxtaposition of "almost" and "absolute nonsense" certainly encourages it.

I've got a lot of background advantages, though; I regularly read technical papers in disciplines I don't even study. This... gives one amazing faith through experience that one can just skip the parts that don't work, grab the parts one can, and look up one or two words later. Tracking down acronyms and abbreviations is often the hardest part, but occasionally, "skip over technical language and look up that word later" comes up.

This... makes "giving up" not a required response to "what the hell is that," by experience. I can usually read myself in to a new technical field in 2-8 hours, just from its published literature.

Here, of course, it let me pick up the parts that apply, which create a reasonably consistent theme to the thing, in 2-3 readthroughs. Doing this pretty much made the parts that do not apply stand out (as well as some of the grammatical nonsense strings), which basically bounced it on a Turing. When you've got the chunk that (accidentally) makes sense down, contrast sections, well... contrast. Quite a bit. While a quick skimthrough yields the exact response you suggest.

Now, for your case study: prove that this response to your post was either by a human or by a machine.

As prior noted, there is no proof. However, neither are there ingrammatical nonsense strings, nor is there contrast text to any accidentally-cohesive themes.

Seriously, what did you feed that thing for source text? lol...

Anyway, no. there is no real meaningful discourse in totality. It's just gibberish. But, so too, keep in mind that my methodology was focused on creating something that a cursory glance might... might... pass the Turing.

...and like you've said, it will bounce most people.

Ever consider using that for the cause? I mean, that and a wordpress is a quick academic journal. Most folks will just glaze over, it ends up completely innocuous, and, well, there are ways to exploit this, from page-ranking manipulation juice to burying equally-obscure but human-written tracts in the "reputable journals." You could even have a chance to make money (or crypto at least) doing so, and hey... your computer-generated poetry is a bit better than mine.

I know, for a fact, it did pass a college TA's cursory glance.

Crap like this has made it into actual journals, because that's the level of proofreading and peer-review that goes on.

Generally, it was submitted as a test and subversion, because the human-written version of that is freakishly common. I often get really annoyed with "journalism," (as much a computer-generated activism option as academic journaling, though probably needing more human rewrite), because it often, well, is devoid of content. I do not care about my feelings when reading a piece, there's music for that; I care about information.


Point being, both human- and computer-generated forms are... common, actually, even beyond the faking-your-way-through-college and into the professional world. It's become a comedy, been done-and-revealed for the expose value, and the like. So it's more of a systemic issue than a Turing success at the moment; apparently, most professional gatekeepers do not themselves pass a Turing test as readers, despite being human, leading to, well, corruption in the industry.

I'm serious about the computer-generated activism thing. How many of those can it put out in a second, exactly? You could have a hundred, or a thousand, "reputable" journals willing to publish anything we want, with filler content to cover.

Think about it.


Follow ups:

Post a response :

Nickname Password
E-mail (optional)

Link URL (optional)
Link Title (optional)

Add your sigpic?