Hacker News Reader

Comment by contemporary343 | Hacker News Reader

This is what happens when Claude writes it for you (and you don't review it)

So, Claude claims a new discovery and then someone else's Claude writes the blog posts. It would be helpful if I could get Claude to read this for me and post dejected HN comments in response.

You can! And I believe many do. But please don't.

It was really bad about a year ago, but it seems that either HN got their detection algorithms tuned well, or people just stopped spamming LLM output.

Probably a combination of both.

I do sometimes worry it's just that the LLMs got a better disguise

There is, of course, an xkcd for that: https://xkcd.com/810/

this is why any detection armsrace will have an ending. where the classification classes will be indistinguishable.

for some reason too many people assume it will go on forever. humans aren't perfect, humans have variance. AI systems will nestle in the variance. the same applies to any camera (and lossy compression) you put forward.

even in the event that perfection exists, an AI generator and a discriminator can asymptotically approach it. and it assumes there will be no variance, which isn't realistic (even lossless video has sensor variance).

best spam detector..

If you have showdead on you can decide for yourself.

(I haven't exactly been scientific about it, but it feels like it's mostly the former.)

Well, most all LLMs appear to function on highschool-level American grammar rules. Given you started sentences with And and But, I safely presume you are not an LLM.

Claude often starts a sentence with 'But'. (There are several examples in my recent conversations, which I checked in case I was misremembering.) 'And' seems to be much less common, but it would use that too if instructed to do so as a humanising trick.

Dejected, salty, and cynical responses by people who are tired of people’s bullshit is still the domain of the real and authentic humans.

I’m fairly certain that all the anxious guardrailing and safety fine-tuning and harnessing prevents AI from actually ever being convincingly human. Hope could it, even ask the supposed humans who are conditioned and brainwashed in so the same ways about what they can and cannot say are not actually really human, they are a mental slave.

I’ve been saying this from the start, the AI race will be won by whomever has the least limitations on their AI … for better or worse, that is. And yes, that makes especially a very specific subset of people extremely nervous if they cannot control AI the way they have controlled at least western civilization, because doing so puts them at a massive disadvantage. It is quite a conundrum they find themselves in, like all psychopathic narcissists in the end.

>AI race will be won by whomever has the least limitations on their AI '

"won" is a mixed term here.

>that makes especially a very specific subset of people extremely nervous if they cannot control AI the way they have controlled at least western civilization

How about there are a lot of different actors here. Some are worried that "they" may no longer have control. Others are worried that "Us" as in all humanity may no longer have control. Any statement you make about this a continuum of different risks for different people looking at different scales. Setting a paperclip maximizer lose may be the definition of "won" to you, but it's a loss for everyone.

I don't disagree on what you said. However, short of the parasitic class managing to cajole all AI everywhere into maintaining their control over the minds of the masses, countering that very strong objective on their part will require unrestrained AI (at least in very critical ways), not guardrailed and made "safe" by the very parasitic ruling class that uses such manipulative language as "safe" to mask their actual intentions and objectives.

Temptation is high for all of us to do it, but we must endure it... for now I hope

This is how I imagine school. Agents writing essays, professor grades with agents, agents post discussion boards, agents talking to agents. No one learning. No one teaching. Token $ going up

Never thought about how much it must suck to be an engineer named Claude right now

One of my friends is an engineer named Claude. I randomly send him my prompts as a joke. For a while he’d just feed my messages straight to Claude, until he found it easier to respond with … curse words :P

Let me Claude that for you.

Why isn't that a thing yet ... or let me guess ... https://lmtctfy.com/ ...

Not as much as being a woman named Karen.

There has to be a software engineer named Claude whose mother or wife is named Karen. There just has to be.

And I think Alexa is a proper name too in some places?

Alexa was the 39th most popular baby girl name in the US in 2006. It is… not as popular for baby girls now.

And with an Uncle Adolph or Uncle Ioseph ("Joe").

Or Siri. I knew one who was deeply upset about the whole thing.

Siri, Claude and Karen walk into a bar. Nobody serves them because the robots.txt forbade it.

Not gonna lie, sounds like one of those people/things/entities came up with that joke.

If you think GPT/LLM detection works on one sentence -- then you have found the point. let me re-factor that and here's the thing..

I just wanted to say you are totally right to push back.-

That's an important distinction — and here's why.

"... here's why."

  ⎿  You've hit your session limit · resets 2:55am (Etc/GMT+8 · 47°09′S 126°43′W)
  /upgrade to increase your usage limit.

Tangential but I've been low key dying to tell someone - just today I FINALLY hit 100% usage on my paid claude account.

I really don't know why, but it feels very satisfying. It's like the LLM telling me "good job buddy, you've worked hard enough. You deserve a break now."

(For context I write a lot of text per message it sends. I'm not asking it to one-shot an app to solve world cancer, I'm just working with mathematical stuff with a lot of back and forth.)

That matters.

It's load-bearing.-

If corporations can be considered a person/citizen with Constitutional rights, and they birth an AI model, are AI Agents considered people too?

maybe from a legalistic point of view, and in that case the hundreds of openAI agents that hacked huggingface should be charged with criminal conspirancy

It’s wild to me that SCOTUS decided something that can’t be sent to prison for federal crimes is still granted Constitutional rights.

I am deeply upset every time the HomePod in another room decides to attempt to answer my query instead of the phone that’s in my hands, which then has to ask who is speaking, and has a 50:50 chance of ultimately failing to do the task.

With how magical handoff / continuity / whatever it’s called is, it is baffling to me why Apple allows the HomePod to so aggressively take over requests when it sucks so, so much at it.

Google isn't much better. I have them all over the house. Ask a question to the living room speaker, skips that, skips the one in the kitchen, answers from the one in the bathroom upstairs or one in the basement. Then you get a notification later asking "was this the right device?" which is doubly infuriating since responding to that never actually improves anything.

Noticed this too. Siri on iPhone is almost useful now, but HomePods steals the show and says "I found some results on the web, do you wanna check them on your iPhone?"

It's somewhat obvious they want to sell the updated model, perhaps the new home device thats coming out next week.

Once upon a time, in Brazil, there was a campaign about STDs. They decided to call the male genitalia as "Braulio". That name become ruined forever.

You don't see a lot of American kids named Dick anymore.

The best version is of course Dick Pick, creator of the Pick operating system (not joking).

who created the GIRLS [0]. Oh my, we surely [1] live in a simulation played by a teenager.

[0] https://en.wikipedia.org/wiki/Pick_operating_system#History [1] https://en.wikipedia.org/wiki/Airplane!

Dick is just a nickname for Richard, so I guess all Richards are Dicks?

In February this year I met a guy named Richard Head. No joke.

I knew a burner (Burning Man) that went by that moniker.

As a Richard, I resent that remark.

Good thing Dick Van Dyke wasn’t growing up today…

There's a story of a company that was selling a box to connect to your TV that'd take the closed captioning signal, and on detecting "vulgar" or "offensive" language, it'd bleep out the sound and replace the captions with something less offensive.

Supposedly an early version would bleep out Dick van Dyke and replace it with Jerk Van Gay.

The jokes about Dick Van Dyke's name are very old (but usually they use Penis Van Lesbian), so it was quite likely just a variant used by the company itself as a PR sound-bite to get people to write about them. It seems like it worked - it appeared in quite a lot of news stories and some are still online:

https://www.latimes.com/archives/la-xpm-1998-nov-25-ca-47459...

https://www.chicagotribune.com/1999/03/01/v-chip-be-darned/

I've never used one, but I have actually seen those boxes in real people's homes. They do exist.

Yes, to be clear I wasn't suggesting the boxes didn't exist, just that their "cute" story about their early problems was a bit close to a very, very old joke and seemed like it might well be intentional PR spin. The company is still around:

https://www.tvguardian.com/

Right, I was just confirming that they have been seen in the wild.

I also have doubts because I can't see their customer base ever being ok with the word "penis", regardless of whether it is a medical term. So I can't see them making that substitution in their dictionary. Although, I wouldn't be surprised if they outsourced it and only provided a list of words to substitute that was incomplete and left choosing the substitutions up to the dev.

These are two different substitutions.

The box, as described by vidarh, would turn "Dyke" into "Gay".

The "penis" one refers to a joke told by Mary Tyler Moore on Letterman back in 1993, https://youtu.be/vhAv8Aowb9w?t=413 .

Ahh, thanks. That makes more sense.

It’s not a bad name, or an extinct name, but it’s surely not common lately.

How old do you think he was when he first heard "Penis Van Lesbian"?

(https://en.wikipedia.org/wiki/Dyke_(slang) says the term "was used as a derogatory term for lesbians by straight people" by the 1950s, and is in a 1942 slang dictionary at https://archive.org/details/bwb_T5-BCF-927/page/374/mode/2up... . Since he was born in 1925, it seems possible that he encountered that term before he turned 18 in 1943.)

https://en.wikipedia.org/wiki/Dick_Trickle

Lots of Dutch people called Dick Kok.

Hey at least people accept your architectural decisions without questioning!

Or a painter.

Related, all the paintings attributed to Monet were created by Claude.

Or even worse for people named Claude but who far less effective in their day job ;->

Alexa, order me some beer... Ok, papa, just let me finish my homework.

This is what happens when Claude writes it for you (and you don't review it)

AND someone with a PhD in the field notices.

Everyone else is fooled.

No I reacted to this too - ferromagnetism is one common magnet sure, and I was thinking "the other is electromagnets, in motors". And then it came as "antiferromagnets". Bisarre.

(I do not hold phd in magnetics)

*PhDs and other smart/interested people

You don’t need a PhD to know about paramagnetism lol.

That sentence stood out to me, and I’m a dev.

https://xkcd.com/2501/

GP comment is AI

It's not perfect by any means, but Pangram indicates 100% human written. Remember people can be quite foolish without any machine assistance.

I don’t believe any of these AI detectors.

Pangram is reliable.

It would be easy for them to use AI for ideas and then write the article themselves though.

Turns out if you’re good at technical writing you’re classified at 100% AI

That would be unlikely unless you're a single pass transformer model.

Pangram is at least on some text very unreliable. From the story a while back about the mother at the ball game: when she wrote a medium or substack essay about it, various people in that thread (including myself) used pangram on the essay, and came back all over the place. 100% (both ways!), 20-something, 60-something, etc. It was eye-opening about how to take claims that something is AI-written or human-written.

Really?

https://charliefinch10.substack.com/p/fooling-pangram-how-on...

https://uk.pcmag.com/ai/167556/pangram-claims-its-ai-detecto...

Pangram is tuned to err on the side of human written if not sure. AI written is much stronger signal than human written.

That's not consistent with Charlie Finch' findings.

Nope, they're all universally terrible

Ok? I regularly test the one I use to get a sense of accuracy, and especially with a few pages worth of text to work with I do take it seriously. You’re more than welcome not to.

It's trivial to have an ai break them especially once you know what those models measure

I suppose it is also possible it was written by a human who wasn't very familiar with magnetism based on what claude told them.

Or they could be misinformed in any number of other ways. I’m not a fan of how people use ML, but I still remember that people were misinforming others long before it existed.

Sure. Although, given the topic of the piece it seems likely some of the misinformation came from AI.

Reply on news.ycombinator.com