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

"Don't You Just Upload It to ChatGPT?"

correresmidestino.com

Post reply on HN
Print view View original

271–280 of 408 posts

Re: "Don't You Just Upload It to ChatGPT?"

#271

Post by Lalabadie »

TZubiri wrote: post #241
Drupon wrote: post #14

An honest to god article full of em dashes that's not because it was AI but because it was a human using them as a crutch to get around crafting sentences that flow naturally. Almost brings a tear to my eye.

Either it's LLM generated, or it's written by someone who wants to be ambiguous about using LLMs. Either way, I'm not reading it, it's a clanker or a clanker collaborationist. I mean, how would you even write an em dash? There's no button in the keyboard for em dashes, it's not in ascii, it's just not something we write in internet text with, it's a safety watermark put into LLMs by OpenAI to help making LLM generate…

  > I mean, how would you even write an em dash?
⌥ ⇧ +

It's been seared into my muscle memory for more than a decade. I keep using it, too. It's present in the popular training sets – and then in LLM outputs – simply because it's proper punctuation.

Re: "Don't You Just Upload It to ChatGPT?"

#272

Post by dijksterhuis »

Kiro wrote: post #217
CGMthrowaway wrote: post #76

Earlier quoted context omitted.

Well said. Everyone agrees AI can't do their job, so it ends up doing everyone else's. I'm not sure how to formulate it yet but it seems there is some Peter Principle/Gell-Mann Effect corollary that is AI-related we can say here. Perhaps: "AI rises to the level of its users' incompetence." Or: "Confidence in AI output is inversely proportional to one's ability to verify it"

> Everyone agrees AI can't do their job, so it ends up doing everyone else's. In real life I haven't met a single programmer who doesn't think AI can do their job. If someone would actually say that I would immediately think they have hubris and overestimate their skills.

[dead]

Re: "Don't You Just Upload It to ChatGPT?"

#273

Post by themafia »

GreenSalem wrote: post #270

I had transliterated lyrics of a song * with stanzas in Urdu , Braj Basha, Persian and Arabic , that I wanted to understand better .. Gemini did a pretty good job of translating this to English . Sure a professional human translator would have done a more nuanced job if I was willing to invest the money and time . But ... * tajdar e haram originally by Payam Saihalwi, later versions by the Sabri Brothers and recently…

Is the assumption that the LLM did the translation? Or that it just understood your query and submitted, on your behalf, to a tool you could have just used directly?

Re: "Don't You Just Upload It to ChatGPT?"

#275

Post by tombert »

lambda wrote: post #268
oytis wrote: post #161

Earlier quoted context omitted.

Try translating some prose from English to another language, then, in a different model, back to English

I tried this with the original comment in the thread. Guaranteed to not be in the corpus, references a few terms that also wouldn't be in the corpus (Claude Fable), and long enough to be more than a sentence or two while short enough to compare in a discussion like this. I did this with entirely local models I have sitting around on my laptop. Minimax M2.7 at a 3 bit quant with 8 bit quantized KV cache for English ->…

This is terrible. I never use em dashes!

Re: "Don't You Just Upload It to ChatGPT?"

#276

Post by phendrenad2 »

Reminded me of this quote:

"Expertise in one field does not carry over into other fields. But experts often think so. The narrower their field of knowledge the more likely they are to think so." - Robert Heinlein

In this case, the gym buddy doesn't think that she's an expert in the other field, but dismisses it as something ChatGPT can do with ease.

Re: "Don't You Just Upload It to ChatGPT?"

#277

Post by shnock »

not_a_bot_4sho wrote: post #196
toddmorey wrote: post #135

Earlier quoted context omitted.

I always imagine the model rolling its silicon eyes when it’s assigned a personality (“you are an expert growth hacker”) at the start of the prompt. Was that ever actually shown to be effective? Is it still?

> Was that ever actually shown to be effective? Is it still? Yes! Personas demonstrated measurable improvement in a few different ways, with caveats of course. The common intuition is that personas influence token space in beneficial ways. I'll come back here later on desktop and link a few (still) relevant papers on this topic.

Please do, thank you! I have been similarly skeptical as your comment's parent

Re: "Don't You Just Upload It to ChatGPT?"

#280

Post by tombert »

lelanthran wrote: post #143
tombert wrote: post #139

Earlier quoted context omitted.

I don't necessarily disagree, but there's gotta be a name for some kind of "infinite extrapolation" fallacy, where you assume that the current rate of progress will continue indefinitely. That might happen, but I don't think it's implied, at least given literally every other bit of technology that has ever happened in history ever.

> I don't necessarily disagree, but there's gotta be a name for some kind of "infinite extrapolation" fallacy, where you assume that the current rate of progress will continue indefinitely. I am not assuming they'll continue indefinitely, but it's a small step from writing code to planning out the code to write, and another small step from planning a coding project to planning a software project, etc. These are all s…

I think with a fixed size problem, no we wouldn't be able to demand the same salaries that we get now.

I dispute that the problem is fixed size. The people who are senior engineers now will learn how to think at a higher level with the AI models.

Post reply on HN