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

AI Detectors Get It Wrong. Writers Are Being Fired Anyway

gizmodo.com

Post reply on HN
Print view View original

151–160 of 189 posts

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

#151

Post by TeMPOraL »

dartos wrote: post #148
shmel wrote: post #129

Earlier quoted context omitted.

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…

> What is the correct text anyway? Exactly. The fact that language is fuzzy is why LLMs work so well. The issue is that most people expect computers to not make mistakes. When you write a formula in an excel sheet, the computer doesn’t mess up the math. The average non tech person knows that humans make mistakes, but are not used to computers making mistakes. Many people, maybe most, would see an answer generated by…

The word you're looking for is "any", not "average". As in, it can write like any human, any way you want it to. Not just as some "average human".

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

#152

Post by zamadatix »

amonith wrote: post #90
retrac98 wrote: post #70

Earlier quoted context omitted.

But how do you know we’re not there yet? Not across the board, but isn’t it possible there’s a small yet growing portion of written content online that’s AI generated with no obvious tells?

I think we have a misunderstanding - I don't mind if I'm reading AI generated content as long as it doesn't look like "the typical AI content" (or SEO slop). In my point of view companies/writers might use AI detectors to continue improving the quality of their content (even if it's written by hand, those false positives might be a good thing). We're not there yet because I still see and read a lot of AI/SEO slop. I…

I don't follow how you're reaching your conclusion. You only mind reading AI content when it's obviously AI/slop and you conclude the vast majority of decent content is not AI generated. In your conclusion how were you able to identify good content as being written by AI or not?

E.g. it's perfectly possible that in terms of prevalence "AI slop > AI acceptable > human acceptable" instead "AI slop > human acceptable > AI acceptable" and nothing noted explains why it is one instead of the other.

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

#153

Post by CyberDildonics »

perihelions wrote: post #150
CyberDildonics wrote: post #141

Earlier quoted context omitted.

They were talking about logging your own encrypted keystrokes and being in control of them. This would be dehumanizing? This means 'hacker news has lost their way'? Logging your own keystrokes and encrypting it is 'bleakness of tech'? This is a 'surveillance machine'? What are you talking about?

If you feel compelled to surveil yourself so as not to be arbitrarily fired by an algorithm, I do consider that dystopian; yes. You're not "in control" of data you're expected to turn over to your employer to keep your job. Worse still if these keyloggers become normalized, and they'll shift from being "optional" to "professionally expected" to "mandated". This (IMHO) is an example of an attempt at a technical soluti…

surveil yourself

Surveillance is something you do to someone else. If it's yourself you're just keeping records. It's common that proving validity of something involves the records of it's creation. Is registering for copyright surveillance?

data you're expected to turn over to your employer

If you got paid to make something, that would be your employer's data anyway.

Worse still if these keyloggers become normalized, and they'll shift from being "optional" to "professionally expected" to "mandated"

You think a brainstorm about using a blockchain by a hacker news comment is going to suddenly become 'mandated'?

And in several EU countries it's illegal to have surveillance cameras pointed at workers without an exceptional reason

They described logging their own keystrokes and encrypting them to have control over them. It isn't a camera and it isn't controlled by someone else. Also they said in an editor, so it isn't every keystroke, it would only be the keystrokes from programming.

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

#154

Post by Filligree »

NoMoreNicksLeft wrote: post #149
chipotle_coyote wrote: post #144

Earlier quoted context omitted.

> Also, has it not occurred to anyone that deep down in the brainmeat, humans might actually be employing some sort of organic LLM when they engage in writing? This is a fairly common take, along with the idea that AI image generators are just doing what humans do when they "learn from examples". But I strongly believe it's a fallacy. What generative AI does is analagous to what humans do, but it's still just an anal…

> But I strongly believe it's a fallacy. Either the LLM is the most efficient way to generate text, or there's some magic algorithm out there that evolution stumbled upon a million years ago that we haven't even managed to see a hint that it exists. In which case, you'd be right, this is a fallacy. Or, brainmeat can't do it better or more efficiently, and either uses the same techniques or something even worse. The l…

> Or, brainmeat can't do it better or more efficiently, and either uses the same techniques or something even worse. The latter seems unlikely, humans still do pretty well at generating text (gold standard, even).

Considering we use something like a thousand times the compute, "something even worse" seems plausible enough.

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

#155

Post by shmel »

dartos wrote: post #148
shmel wrote: post #129

Earlier quoted context omitted.

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…

> What is the correct text anyway? Exactly. The fact that language is fuzzy is why LLMs work so well. The issue is that most people expect computers to not make mistakes. When you write a formula in an excel sheet, the computer doesn’t mess up the math. The average non tech person knows that humans make mistakes, but are not used to computers making mistakes. Many people, maybe most, would see an answer generated by…

What do you mean they aren't used to computers making mistakes? Have they ever asked Siri/Alexa something and got useless answers? Have they ever seen ASR or OCR software making mistakes? Have they called semi-automated call centers with prompt "say what you need instead of clicking numbers" only to hear repeated "sorry, I don't understand you" until you scream "connect me to a bloody human"? Have they ever seen a situation when automated border control gates just don't work for whatever reason and there are humans around to sort this out? Have they ever used google translate in last 20 years for anything remotely complicated, like a newspaper article? Have they ever used computers for actual math? Is computer particularly good at solving partial differential equations, for example? Have they ever been in a situation where GPS led them to a closed road or a huge traffic jam? Have they ever played video games where computers sometimes make stupid things?

Sure, computers are better at arithmetic humans, but let's be honest, nobody uses chatgpt as a calculator. Last 20 years AI is getting everywhere, we keep laughing that sometimes AI systems make very obvious stupid mistakes. Now we finally have a system that makes subtle mistakes very confidently and suddenly people are like "I thought computers are never wrong". I can't fathom how anyone would expect that.

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

#156

Post by nerdponx »

boredpudding wrote: post #146
nerdponx wrote: post #108

Earlier quoted context omitted.

The flipside is that the AI-written code I've seen at work is usually painfully obvious upon human code review. If you need a tool to detect it, either it's good AI-written code, or you have particularly inept code reviewers.

Be careful here about confirmation bias. If you only spot 10% of the AI-written code, you'll still think you see all of it, because a 100% of the ones you spot are indeed AI-written. And the 10% you see, will indeed be painfully obvious. The ones you don't notice aren't obvious.

That's fair.

It depends on why you care about AI-written code.

At the code review stage, we care mostly that the code is good (correct, readable, etc). So if the AI-written code passes muster there, then there's nothing wrong with it being "AI-written" in our eyes.

If you care about AI-written for the sake of preventing AI usage by your developers, then I think it's already impossible to detect and prevent.

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

#157

Post by luma »

jgalt212 wrote: post #64
luma wrote: post #61

Earlier quoted context omitted.

OpenAI announced that they had started on an AI text detector and then gave up as the problem appears to be unsolvable. The machine creates statistically probable text from the input, applying statistics to the generated result will show nothing more than exactly that. You’re then left triggering false positives on text that is the most likely which makes the whole thing useless.

> OpenAI announced that they had started on an AI text detector and then gave up as the problem appears to be unsolvable. Making a reliable LLM also appears to be unsolvable, but we still work at it and still use the current wonky iterations. My comment is even if there is no perfect AI detectors, a lot of these tools are good enough for a "first pass"--coincidentally the same use case many effective LLM practitioner…

Sure it could maybe be kinda right, but what is the cost of a false positive? If you have, say, a 10% false positive rate, and there are theoretical reasons to think you’ll never get that anywhere close to zero, then what use case does this serve? Hey student, there’s 90% chance you cheated, well no I’m that 10%. What now?

Again, OAI cancelled work on this believing it not to be solvable with a high degree of confidence. What is the use case for a low confidence AI detector?

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

#158

Post by immibis »

immibis wrote: post #49
squircle wrote: post #33

This content may violate our usage policies. Shit. Let me try this again... On the plus side, once humanity completely loses faith in the perceived value of AI, we will no doubt (I hope) wake up and realize the value of true human connection--unplug ourselves from this awful beast, and begin to rediscover what it means to be human.

"I'm sorry, but I am unable to comply with that request": https://suno.com/song/50291ff1-f971-4bff-b10e-fd168161cf3c

50% chance to get the one with audible lyrics, and I picked the wrong one. Suno always generates two songs from each prompt. Here's the same thing but you can actually hear it. https://suno.com/song/1f243ba3-f64d-4cac-b4d8-4f7aa1a64fcc

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

#159

Post by QuantumGood »

neilv wrote: post #48

> Mark sent his client a copy of the Google Doc where he drafted the article, which included timestamps that demonstrated he wrote the document by hand. It wasn’t enough. Mark’s relationship with the writing platform fell apart. He said losing the job cost him 90% of his income. The article is a little vague, but, assuming Mark is telling the truth, and the article is reporting reasonably, then I can think of a few p…

Or a more specific subset: The boss's boss: "Can't we just do that with AI? But don't let anyone know that is our intention."

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

#160

Post by anigbrowl »

meindnoch wrote: post #88
TeMPOraL wrote: post #72

Earlier quoted context omitted.

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…

>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 physical keyboard according to LLM output. That's how the first automated trading firms operated in the 80s. NASDAQ required all trades to be input via physical terminals, so they build an upside down "keyboard" with linear actuators in place of the…

That article is infuriating. Classic 'it was great when I did it, now it's gone too far.'
Post reply on HN