Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Fake Books

lcamtuf.substack.com

Post reply on HN
Print view View original

171–180 of 371 posts

Re: Fake Books

#171

Post by jefftk »

thom wrote: post #156
glenstein wrote: post #152

Earlier quoted context omitted.

I don't understand how you think this is responsive to the issue of fabrication. The article already considered the possibility of the human content farm and concluded it was unlikely on the basis of information that was completely fabricated. The fact that the commenter proceeded to generalize, show a link, etc is not responsive to that. A responsive answer would be along the lines of "well, a content farm might fab…

"If you find that interesting, you might find this interesting" seems like a perfectly reasonable motivation for an HN comment. I think it's worth forgiving that and enjoying the broader horizons rather than treating HN threads as a way for lots of smart people to try and craft the most definitive statement about one narrow issue, while policing any deviation therefrom.

> "If you find that interesting, you might find this interesting" seems like a perfectly reasonable motivation for an HN comment.

But that's not the form of the comment. It begins with "I think there’s a good chance it was ghost written by a content farm"!

Re: Fake Books

#172

Post by thom »

jefftk wrote: post #171
thom wrote: post #156

Earlier quoted context omitted.

"If you find that interesting, you might find this interesting" seems like a perfectly reasonable motivation for an HN comment. I think it's worth forgiving that and enjoying the broader horizons rather than treating HN threads as a way for lots of smart people to try and craft the most definitive statement about one narrow issue, while policing any deviation therefrom.

> "If you find that interesting, you might find this interesting" seems like a perfectly reasonable motivation for an HN comment. But that's not the form of the comment. It begins with "I think there’s a good chance it was ghost written by a content farm"!

If I were a content farm I'd probably use text cut and pasted from text models, but I'd never heard of the Mikkelson Twins and it was interesting to learn about them.

Re: Fake Books

#173

Post by zefhous »

A couple years ago we ordered an "on-demand" printed book from Amazon for an old public domain book. It came misprinted, with a book on "Mobile Solar Power" right in the middle of it.

I took a look through it cause I thought it would actually interesting, and found it to be this type of fake book that looks to be instructional on the surface, but in reality is not actually educational in the least, and in fact dead wrong in many places.

I posted a couple excerpts on Twitter here, 3-tweet thread. https://twitter.com/zefhous/status/1281679975339274241

Re: Fake Books

#174

Post by kmeisthax »

smilespray wrote: post #43
planetsprite wrote: post #36

Earlier quoted context omitted.

I get the feeling it was outsourced to a content farm, which outsourced itself to GPT-3

...which then outsourced itself to GPT-2...

Given the way that OpenAI sources their training corpus and the amount of people using GPT-3 I would not be surprised if GPT-4 winds up getting trained on a large amount of GPT-3 output.

Think about it - the best niche GPT-3 has is generating plausible spam. If you just need a lot of text, but you don't care about what it says[0], you're going to write it using the cheapest possible tool. OpenAI's training corpus is sourced through web crawls, so all of that spam is destined to get recycled back into the next generation of GPT.

[0] For example, if you want to be able to post a bunch of political spam that looks like genuine comments on a web forum. See GPT-4chan[1] as a practical example of this.

[1] A tweaked version of GPT-3 using 4chan's politics board as training corpus.

Re: Fake Books

#175

Post by dEnigma »

Contrary to what the article says it IS possible to find information about NFP chemotherapy on the internet. E.g.:

  Of these patients, 44 were treated with HAIC and 20 with sorafenib. HAIC involved cisplatin (50 mg fine powder in 5-10 ml lipiodol) and a continuous infusion of 5-fluorouracil (FU) (1,500 mg/5 days), which is referred to as new 5-FU and cisplatin therapy (NFP). [1]
[1] https://pubmed.ncbi.nlm.nih.gov/29285366/

Re: Fake Books

#176

Post by dEnigma »

moss2 wrote: post #92
TaylorAlexander wrote: post #16

I think there’s a good chance it was ghost written by a content farm for someone who wanted to publish a book in a popular category to make a quick buck. I recently learned that this is a soft-scam technique advocated for by a couple of guys called the Mikkelson Twins. They say you can make money by publishing audiobooks on audible by finding a popular topic on audible, then hiring a ghost writing service to write yo…

> I think there’s a good chance it was ghost written by a content farm for someone who wanted to publish a book in a popular category to make a quick buck. The article literally says > My first theory was that the writing must have been outsourced to a content farm, where multiple individuals worked on portions of the text without caring to understand the context. But the explanation had a flaw: there’s nothing you c…

Only that it's indeed possible to find information about NFP chemotherapy online, e.g. here https://pubmed.ncbi.nlm.nih.gov/29285366/

  Of these patients, 44 were treated with HAIC and 20 with sorafenib. HAIC involved cisplatin (50 mg fine powder in 5-10 ml lipiodol) and a continuous infusion of 5-fluorouracil (FU) (1,500 mg/5 days), which is referred to as new 5-FU and cisplatin therapy (NFP).

Re: Fake Books

#178

Post by hoistbypetard »

nicbou wrote: post #118
derefr wrote: post #46

Earlier quoted context omitted.

> Except what the Mikkelson Twins actually make money on is selling you on a “money making course” where they tell you to do the above. So there’s all kinds of people trying to follow this method out there, it seems. How do people still become wealthy selling "money making courses", when enough people have created "money making courses" that they are, themselves, a commodity?

You sell a money making course course generator and sell that

Money making course course generator as a service. MMCGGaaS. It's the next big thing.

Re: Fake Books

#179

Post by phtrivier »

"90% of everything is crap" ; but, now, 'everything' is computer-generated, and therefore, infinite. I'll let mathematicians study the convergence, while physicists conclude that "everything is crap", give up and get a beer, and software engineers go back to fixing the other 90%.

Re: Fake Books

#180

Post by DoneWithAllThat »

I find it to be pretty obviously ML generated. As in even the parts about NFTs are written in a style that is patently obviously not authored by a human. I’m sure the difference will become less noticeable over time but for now AI-authored prose is easy to spot.
Post reply on HN