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

The revolt of the reader

bcantrill.dtrace.org

Post reply on HN
Print view View original

281–290 of 311 posts

Re: The revolt of the reader

#281

Post by pessimizer »

jbeninger wrote: post #169
somat wrote: post #145

Earlier quoted context omitted.

Very depressing, and I believe it. I am a bit of a luddite in this domain and have so far managed to resist the lure of using the generator to expand my thoughts, and I still catch myself writing "it's not just an X it's a Y" and other generator type tells. If it infecting my patterns it is totally entering the wider subconscious as "How to write" (Sighs)

That's not a generator tell when used judiciously. It's a genuinely useful construction that's been poisoned by overgeneration.

It's bad in and of itself the way LLMs use it. They separate it into a weird declaration when it's usually something you'd say at the beginning and/or end of a paragraph or passage where you're making a case for that statement.

It's part of how LLMs often seem to repeat the same thing over and over again. Most arguments are X is Y (and the others are X is not Y, or X is likely Y), and LLMs lean on "X is Y" because it is the most distilled form of that argument - a thesis minus the meat. Also "This is not a Y, this is a Z."

They're just thesis sentences inserted in inappropriate locations. The reason it makes them overdramatic is because separated from the argument, they seem like they should be premises. Conclusions without premises have the feeling of demands.

Re: The revolt of the reader

#282

Post by cgio »

sersi wrote: post #235
OroPla wrote: post #166

Earlier quoted context omitted.

I already notice LLM speech patterns in people that use them a whole lot. If you speak more than one language I recommend talking to LLMs in a language that you don't use when speaking to people.

I have a tendency to switch between different languages when interacting with LLM depending on the language of the most likely sources the LLM will pull. It's interesting because the llm speech patterns vary from language to language. And I am much faster at noticing LLM generated text in English (which I use 70% of the time) than in French or Spanish.

In Greek they are still atrociously bad, still feels like they translate an English thought. Nevertheless, the etymological insights they can give on words or Ancient Greek passages is making reading the ancients really fun.

Re: The revolt of the reader

#283

Post by emil-lp »

droidjj wrote: post #280
runako wrote: post #73

Disclaimer: I do not like to read LLM-generated text any more than anyone else. IMHO a big problem with Pangram in particular is that they market it as a reliable tool that can be used to catch students cheating. This can obviously have disastrous effects on young lives, because it is not as reliable as they suggest. Per their own benchmarks, they do not achieve 100% accuracy even on text that is published on the Int…

The false positive rate for Pangram 4 is something like one in 24,000.[0] To put that in perspective, the wrongful-conviction (false positive rate) for death-sentenced defendants in the US is estimated conservatively to be around 4.1%.[1] The FP rate for death-sentence convictions is 1,000 times bigger than Pangram’s FP rate. Now, the US criminal system is not a great yardstick for justice. But it goes to show you Pa…

> The false positive rate for Pangram 4 is something like one in 24,000.

First of all: says them.. Do you think they might have an incentive to boost their numbers?

Second, this is on existing text, which might be in their training data, no?

Re: The revolt of the reader

#284

Post by matheusmoreira »

> Or, better yet, consider doing what generations of writers have done before you, and treating that prompt as a skeleton that you use to write your piece yourself!

Pangram complains about "paraphrased" content though. Sigh...

Re: The revolt of the reader

#285

Post by matheusmoreira »

kccqzy wrote: post #35

I would love to use Pangram but they simply don’t allow signing up with my custom email domain. The error was “This email address can't be used for signup. Please use a different email.” I’m not about to create a Gmail is to use your service. To me the attack on the decentralized nature on Internet infrastructure is no less serious than the attack on the human provenance of writing itself.

For the record, I was able to sign up using my matheusmoreira.com domain.

Re: The revolt of the reader

#286

Post by lucb1e »

droidjj wrote: post #280
runako wrote: post #73

Disclaimer: I do not like to read LLM-generated text any more than anyone else. IMHO a big problem with Pangram in particular is that they market it as a reliable tool that can be used to catch students cheating. This can obviously have disastrous effects on young lives, because it is not as reliable as they suggest. Per their own benchmarks, they do not achieve 100% accuracy even on text that is published on the Int…

The false positive rate for Pangram 4 is something like one in 24,000.[0] To put that in perspective, the wrongful-conviction (false positive rate) for death-sentenced defendants in the US is estimated conservatively to be around 4.1%.[1] The FP rate for death-sentence convictions is 1,000 times bigger than Pangram’s FP rate. Now, the US criminal system is not a great yardstick for justice. But it goes to show you Pa…

> The false positive rate for Pangram 4 is something like one in 24,000.

Gotta suck to be one of the 8B/24k=~300k people in the world whose writing pattern is falsely labelled as slop by this tool that people say is so accurate so customers are going to feel really sure about your alleged dishonesty about writing your own texts

This false positive rate is a double-edged sword. Please still be careful when accusing people

Re: The revolt of the reader

#288

Post by matheusmoreira »

somat wrote: post #143
kierangill wrote: post #65

My revolt is against the cognitive stress of reading generated text. A trope typically indicates I’m in for an uphill read. I recently read this William Zinsser quote that inspired a nickname for this: Clotted Claude [1]. > Nobody has made the point better than George Orwell in his translation into modern bureaucratic fuzz of this famous verse from Ecclesiastes: > > I returned and saw under the sun, that the race is…

My take is that the generated stuff is terrible for communications. It feels great to use, direct your machine minion to fill out your thoughts for you, but holy hell does it suck to be on the receiving end. Least of all is the disrespect, they don't care enough to even talk to you but worse is having to try and reason through that big incoherent blob. Probably to only reasonable thing to do is to try and get your ow…

I used to sympathize with those on the "receiving end" but I'm becoming extremely desensitized to their plight.

The sheer amount of hatred they level towards anyone who's using AI or is even just positive towards it is something to behold. They've got racism-tier slurs for us now.

The negativity and stigma are so strong I've stopped caring. I don't feel like putting in human effort for the benefit of these hateful people anymore.

Re: The revolt of the reader

#289

Post by AnimalMuppet »

apparent wrote: post #272
AnimalMuppet wrote: post #267

Earlier quoted context omitted.

I'm not absolutely opposed to AI in principle. If it writes things worth reading, I'm not opposed to reading them. The problem is when it writes things that have the appearance of making sense but actually don't, when it writes what could be better put in 10% of the words, when it writes with no coherent logic. I don't have the time or the mental energy to wade through that on the faint chance that there might be a g…

I agree completely, but how does that relate to my comment (saying it's cat and mouse, the game may never end, and Pangram's current success isn't indicative of future value)?

It's saying I don't actually care. I don't care if Pangram can detect AI writing in the future. I don't care if it can't. I'm saying the question does not matter to me.

I don't care about not reading AI. I care about not reading slop.

But I guess I could care about one outcome of this game. If Pangram loses the ability to detect AI writing, but AI writing continues to be this nonsense that has no real point or logic to it, and so more garbage gets published that I want to not read, yeah, I guess I would care about that.

Re: The revolt of the reader

#290

Post by rmunn »

jimmaswell wrote: post #255
rmunn wrote: post #199

Earlier quoted context omitted.

You're running into a cultural difference. The book of Ecclesiastes was written in Hebrew, in a poetical style even though large chunks of it are prose. But if you read other parts of the Bible, such as the book of Psalms (which is entirely poetry, specifically songs, though in many cases we do not know the tune that they were set to), you'll see that Hebrew poetry relied on repetition. For example, here's the King J…

Repetition can make a good impact, but it's beyond excessive in this instance. Contrast with this poem: Non est salvatori salvator, neque defensori dominus, nec pater nec mater, nihil supernum. Translated: No rescuer hath the rescuer. No Lord hath the champion, no mother and no father, only nothingness above. - Eliezer Yudkowsky I find this one profound, and it's given me a lot to think about over the years - in situ…

I never had formal training in Latin, just what I've picked up. But grammatically, I would think the second line would translate to "No champion hath the lord", not "no lord hath the champion". Am I misunderstanding the grammatical suffixes here?
Post reply on HN