Live data from Hacker News

AI Detectors Get It Wrong. Writers Are Being Fired Anyway

gizmodo.com

121–130 of 189 posts

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#121

Startup idea: online text editor that logs every keystroke and blockchains a hash of all logs every day. If you're accused of AI use, you can pull up the whole painstaking writing process and prove it's real.

Stupid startup killing idea: an open-source script that runs LLM in the background and streams its output as input events, so the idiotic keylogger thinks it's all written by hand. Just writing this down here instantly invalidates the premise. An overkill variant to rub salt in the wounds of duped investors: make the script control a finger bot on an X/Y harness, so it literally presses the physical keys of a physica…

> make the script control a finger bot on an X/Y harness,

Too many points of mechanical failure. Just use a RPi Pico W (or other USB HID capable microcontroller) to emulate a keyboard and have it stream key codes at a human pace. Make it wifi or bluetooth enabled to stream key codes from another computer and no trace of an LLM would ever be on the target system.

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#122

Startup idea: online text editor that logs every keystroke and blockchains a hash of all logs every day. If you're accused of AI use, you can pull up the whole painstaking writing process and prove it's real.

I mean, you have both "blockchain" and "AI" in your startup idea. VC money can't be far away :)

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#123

One popular AI detection package that you can licence with the turnitin academic anti-plagiarism software warns that it may produce false positives if the writing is (1) not by a native English speaker, (2) writing on a technical topic, or (3) neurodiverse. So yeah ... congrats, you've built a tool to detect autistic Chinese computer scientists!

Hot takes on American Idol contestants will be free from AI generation, but research papers won't be.

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#124

This is the dawn of a "post non-narrative prose" era. Why bother with an article describing some event or really anything that is traditionally written about in any form of article at all? In time, just the data will be released and your favorite LLM prepared with your favorite reading style will provide the mustard for your data, making it consumable. The journalism industry has failed in their mission, and nobody t…

That's the spirit! Why bother? It's about "consumability". Once meaning and purpose are forever vanquished from life, imagine how happy we will all be drooling through our soulless data stream!

> Once meaning and purpose are forever vanquished from life

Once you truly appreciate the impending death of yourself and everyone you love, and ultimately the heat death of the universe, how do you find meaning and purpose? How meaningful can it be if it's all going away regardless? Being quite morose and consumed by a debilitating sense of pointlessness, it would be really nice to find some hopeful inspiration.

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#125
post #60

Regardless of whether AI detectors work or not, I worry we're reaching a new unfortunate era. While alarmists were worrying about alignment or evil AIs, the true downside of LLMs is turning out to be... AI spam. I've started seeing it everywhere: Q&A sites, forums, etc. In some places it's banned, but how do you reliably detect it? And what makes people post AI spam when there's no reward or money involved? What are…

https://www.nytimes.com/2024/06/11/style/ai-search-slop.html

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#126
post #116
post #83

Earlier quoted context omitted.

You can't regenerate the book from the ISBN. But you can generate the text from the prompt.

Can you, though? I thought LLMs just by virtue of how they work, are non-deterministic. Let alone if new data is added to the LLM, further retraining happens, etc. Is it possible to get the same output, 1:1, from the same prompt, reliably?

They are assuming a lot of things, like the LLM doesn't change, and that you have full control over the randomness . This might be possible if you are running the LLM locally.

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#127

In some cases an AI will make a weird word choice. So do a lot of humans. Sometimes AIs are needlessly wordy. Um...so are a lot of humans. Rinse and repeat. AI detectors are useless. The AIs are training on human writing, so they write fundamentally like humans. How is this not obvious?

But also maybe firing writers who make weird word choices and are needlessly wordy is fine.

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#129
post #107
post #99

Earlier quoted context omitted.

Yeah, I am shocked people keep repeating this. We've seen LLMs easily writing mathematical proofs in the style of Shakespeare, come on.

Saying they find some “average” is an easy way to explain to a layman that LLMs are statistically based and are guessing and not actually spitting out correct text as you would expect from most other computer programs. That’s why it’s repeated. It’s kind of correct if you squint and it’s easy to understand

What is the correct text anyway? Everything around you is somewhat wrong. Textbooks (statistically all of them) contain errors, scientific papers sometimes contain handwavy bullshit and in rare cases even outright falsified data, human experts can be guessing as well and they are wrong every now and then, programs (again pretty much all of them) contain bugs. It is just the reality.

Even very simple ones may require you to twist the definition of "correctness". I open a REPL and type "1/3.0*3.0" and get "0.9999999999". Then you have to do mental gymnastics like "actually it is a correct answer because arithmetic in computers is implemented not like you'd expect".

Re: AI Detectors Get It Wrong. Writers Are Being Fired Anyway

#130
post #99

Earlier quoted context omitted.

Yeah, I am shocked people keep repeating this. We've seen LLMs easily writing mathematical proofs in the style of Shakespeare, come on.

There's still some typicality defined by the prompt though. If you ask for a proof in the style of Shakespeare, you're going to get some "average" Shakespeare. It's kind of embedded in the task definition; you're shifting the reference distribution. If a LLM returned something really unusual for Shakespeare when you didn't ask for it, you'd say it's not performing well. Maybe that's tautological but I think it's what…

> you're going to get some "average" Shakespeare

The point was that no, you wont (necessarily) get some "average" shakespeare. A sampler may introduce bias and look for the "above average" shakespeare in the distribution.

Post reply on HN