Do AI detectors work? Students face false cheating accusations • Hacker News Board

Re: Do AI detectors work? Students face false cheating accusations

#131

Post by Etheryte » Mon, Oct 21, 2024, 7:15 AM UTC

ClassyJacket wrote: post #126

greatartiste wrote: post #111

For a human who deals with student work or reads job applications spotting AI generated work quickly becomes trivially easy. Text seems to use the same general framework (although words are swapped around) also we see what I call 'word of the week' where whichever 'AI' engine seems to get hung up on a particular English word which is often an unusual one and uses it at every opportunity. It isn't long before you real…

How are you verifying you're correct? How do you know you're not finding false positives?

Have you tried reading AI-generated code? Most of the time it's painfully obvious, so long as the snippet isn't short and trivial.

Re: Do AI detectors work? Students face false cheating accusations

#132

Post by CalRobert » Mon, Oct 21, 2024, 7:15 AM UTC

Gud wrote: post #116

llm_trw wrote: post #84

Earlier quoted context omitted.

You can't eat respect.

They are well compensated. https://www.swissinfo.ch/eng/society/swiss-salaries-teachers...

That article, after a very pushy illegal gdpr consent banner, says pay is stagnant and hours long

Re: Do AI detectors work? Students face false cheating accusations

#133

Post by weinzierl » Mon, Oct 21, 2024, 7:16 AM UTC

We had a time when CGI took off, where everything was too polished and shiny and everyone found it uncanny. That started a whole industry to produce virtual wear, tear, dust, grit and dirt.

I wager we will soon see the same for text. Automatic insertion of the right amount of believable mistakes will become a thing.

Re: Do AI detectors work? Students face false cheating accusations

#134

Post by greyadept » Mon, Oct 21, 2024, 7:21 AM UTC

woernsn wrote: post #118

greyadept wrote: post #100

I'd be really interested to run AI detectors on essays from years before the ChatGPT era, just to see if anything gets flagged.

Yes, 3 out of 500 essays were flagged as 100% AI generated. There is a paragraph in the linked article about it.

And another 9 flagged as partially AI.

Re: Do AI detectors work? Students face false cheating accusations

#135

Post by Daz1 » Mon, Oct 21, 2024, 7:25 AM UTC

kreyenborgi wrote: post #108

Daz1 wrote: post #71

Earlier quoted context omitted.

>I thought it was common knowledge that it was impossible to tell whether text was generated by AI. Anyone who's been around AI generated content for more than five minutes can tell you what's legitimate and what isn't. For example this: https://www.maersk.com/logistics-explained/transportation-an... is obviously an AI article.

Obviously false, as LLMs parrot what they're trained on. Not that hard to get them to regurgitate Shakespeare or what have you.

Sounds like a skill issue on your part

Re: Do AI detectors work? Students face false cheating accusations

#136

Post by MathMonkeyMan » Mon, Oct 21, 2024, 7:26 AM UTC

closewith wrote: post #122

prepend wrote: post #22

My kids’ school added a new weapons scanner as kids walk in the door. It’s powered by “AI.” They trust the AI quite a bit. However, the AI identifies the school issued Lenovo laptops as weapons. So every kid was flagged. Rather than stopping using such a stupid tool, they just have the kids remove their laptops before going through the scanner. I expect not smart enough people are buying “AI” products and trusting th…

Reading this comment, it sounds to me that you live in a dystopian nightmare.

Many schools are prisons, same as ever.

Re: Do AI detectors work? Students face false cheating accusations

#137

Post by Ekaros » Mon, Oct 21, 2024, 7:27 AM UTC

anonzzzies wrote: post #124

I don't know what these 'students' are doing, but it's not very hard to prompt a system into not using the easily detectable 'ai generated' language at all. Also adding in some spelling errors and uncapping some words (like ai above here) makes it more realistic. But just adding an example of how you write and telling it to keep your vocabulary and writing some python to post process it makes it impossible to detect…

Sounds like almost same level of effort than actually just writing it yourself. Or getting AI write draft and then just rewriting it quickly. Humans are lazy, students especially so.

Re: Do AI detectors work? Students face false cheating accusations

#138

Post by hau » Mon, Oct 21, 2024, 7:29 AM UTC

acchow wrote: post #129

greatartiste wrote: post #111

For a human who deals with student work or reads job applications spotting AI generated work quickly becomes trivially easy. Text seems to use the same general framework (although words are swapped around) also we see what I call 'word of the week' where whichever 'AI' engine seems to get hung up on a particular English word which is often an unusual one and uses it at every opportunity. It isn't long before you real…

> For a human who deals with student work or reads job applications spotting AI generated work quickly becomes trivially easy. Text seems to use the same general framework (although words are swapped around) also we see what I call 'word of the week' Easy to catch people that aren't trying in the slightest not to get caught, right? I could instead feed a corpus of my own writing to ChatGPT and ask it to write in my s…

I don't believe it's possible at all if any effort is made beyond prompting chat-like interfaces to "generate X". Given a hand crafted corpus of text even current llms could produce perfect style transfer for a generated continuation. If someone believes it's trivially easy to detect, then they absolutely have no idea what they are dealing with.

I assume most people would make least amount of effort and simply prompt chat interface to produce some text, such text is rather detectable. I would like to see some experiments even for this type of detection though.

Re: Do AI detectors work? Students face false cheating accusations

#139

Post by wiz21c » Mon, Oct 21, 2024, 7:34 AM UTC

tightbookkeeper wrote: post #81

gradus_ad wrote: post #12

Seems like the easy fix here is move all evaluation in-class. Are schools really that reliant on internet/computer based assignments? Actually, this could be a great opportunity to dial back unnecessary and wasteful edu-tech creep.

Yep. The solutions which actually benefit education are never expensive, but require higher quality teachers with less centralized control: - placing less emphasis on numerical grades to disincentive cheating (hard to measure success) - open response written questions (harder to teach, harder to grade) - reading books (hard to determine if students actually did it) - proof based math (hard to teach) Instead we keep i…

totally agree. More time spent questionning the students about their work would make AI detection useless...

but somehow, we don't trust teacher anymore. Those in power want to check that the teacher actually makes his job so they want to see wome written, reviewable proof... So the grades are there both to control the student and the teacher. WWW (What a wonderful world).

Re: Do AI detectors work? Students face false cheating accusations

#140

Post by hyperbrainer » Mon, Oct 21, 2024, 7:35 AM UTC

arkh wrote: post #114

add-sub-mul-div wrote: post #40

Earlier quoted context omitted.

Imagine how little common knowledge there will be one or two generations down the road after people decide they no longer need general thinking skills, just as they've already decided calculators free them from having to care about arithmetic skills.

We don't learn directions now: we use GPS. We don't do calculations: computers do it for us. We don't accumulate knowledge: we trust Google to give us the information when needed. Everything in a small package everyone can wear all day long. We're at the second step of transhumanism.

At least the first 2 are far more accurate than humans ever could be. The third, i.e. trusting others to vet and find the correct information, is the problem.