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

Don't paste the AI, please

dontpastetheai.com

Post reply on HN
Print view View original

151–160 of 632 posts

Re: Don't paste the AI, please

#151

Post by whstl »

coldtea wrote: post #138
nchmy wrote: post #112

Earlier quoted context omitted.

I made some ninja edits, so perhaps you didn't see them. But it is entirely possible that both a person and an Ai could produce writing that is both good or bad, useful or misleading. You can also direct the Ai to write about specific points and, again, it may write somehing good or bad. I provided a link to a chat I had with chatgpt about this very conversation and it shows everything I was saying. The fundamental i…

> I made some ninja edits, so perhaps you didn't see them. But it is entirely possible that both a person and an Ai could produce writing that is both good or bad, useful or misleading. Theoritically yes. In practice: - AI responses I get are walls of written diarrhea (even when they're relevant and take our particular context into account). The one's that don't are from people who write well based on their expertise…

> Theoritically yes. In practice:

I get the feeling this is the whole disconnect. People have high expectations from AI but haven't really been on the receiving end of a massive amount of text/code generated by it.

We recently had someone fired for using AI excessively. Before the firing, whenever anyone on the team sat down to give feedback, the only answer was "but everyone is using AI, we are supposed to use AI, how can it be bad?".

Until their boss sat down and forced them to read the verbal diarrhea and they finally admitted that it was 90% hallucinations and AI-filling-the-blanks.

Re: Don't paste the AI, please

#152

Post by chronid »

nchmy wrote: post #118
lurking_swe wrote: post #98

Earlier quoted context omitted.

> But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is... I think it can be OK if AI is used to polish up one’s message or to help structure it. Basically help with grammar, formatting, and wording. But copy pasting an AIs response to a question? Not at all. People earn a reputation over time. TRUST is earned. An AI cannot earn trust. If you copy paste a…

I don't disagree at all. But it still remains that AI can very much be effectively used either to write first drafts OR refine your own first drafts. Though, I think such a writing partner can only really be useful if you know how to write yourself , which also requires knowing how to think (because writing is really just thinking). So, again, in the end all that really matters is whether you are effectively communic…

> But it still remains that AI can very much be effectively used either to write first drafts OR refine your own first drafts.

There is a world of difference between this and copy-pasting LLM output into slack and email.

You can write the first draft with the LLM, or have the LLM help you do an analysis, look at how code works, and so on. But you need to check for the LLM to be correct and you need to be able to articulate what the LLM found in your own words. If you don't you have just offloaded all the work to someone else: someone else will have to check, verify, investigate what could very well be an hallucination (bad context, random bit flips, whatever the cause doesn't matter). Your effective contribution to the discussion will be negative.

Re: Don't paste the AI, please

#153

Post by nchmy »

croon wrote: post #139
nchmy wrote: post #112

Earlier quoted context omitted.

I made some ninja edits, so perhaps you didn't see them. But it is entirely possible that both a person and an Ai could produce writing that is both good or bad, useful or misleading. You can also direct the Ai to write about specific points and, again, it may write somehing good or bad. I provided a link to a chat I had with chatgpt about this very conversation and it shows everything I was saying. The fundamental i…

> The fundamental issue is whether people are taking responsibility to communicate what needs to be communicated, in an effective manner. While I agree with your overall distinction, I believe the scenario you portray is so unrealistic that you will never encounter it. Yes; if I could wrangle Claude into communicating exactly as I would have, the practical difference is of course not observable. But that's never happ…

> if I could wrangle Claude into communicating exactly as I would have, the practical difference is of course not observable. But that's never happened, and I've had an enterprise license at work for years.

Sorry to say, but this is a skill issue (and not necessarily with using Ai). I've used Ai MANY times to help SIGNIFICANTLY improve things that I've written - from a few sentences to many paragraphs.

As I've said in other comments, it'll rapidly and effectively remove all my worst tendencies. And, likewise, sometimes I'll have it write a first draft when I give it appropriate context/guidance that I can't be bothered to fully curate/synthesize. Then, once it gets that draft, I'll take over and tidy it up. You might say that the "can't be bothered" part is the entire issue, and I'll again say that what matters is the final result - a master chef or chief architect is not doing all the grunt work...

Re: Don't paste the AI, please

#155

Post by Hugsbox »

depressedlesor wrote: post #134

What if i have used the AI, it answered, but I don't understand the answer?

If you don't understand the answer, don't answer somebody else with it. You're attaching your name to whatever you copy-paste to people, and if you don't understand it there's a chance it's complete nonsense. Save yourself the embarrassment. If you don't know how to answer somebody's question, just say "I don't know".

Re: Don't paste the AI, please

#156

Post by wccrawford »

seer wrote: post #35

I’m not convinced tbh, a lot of my colleagues messages are “x broken” or “I do y” … no context whatsoever, and then I have to coax their context for making these decisions. Since people started using Claude now I get full context of everything… might be too much sure, but to be honest over-communicating seems better than under-communicating - sure it’s boring and tedious but that shifts the blockage to me. Otherwise…

If they user types "x broken" or "i do y" into Claude or ChatGPT with no context, that's absolutely no better than just sending it to you.

I'd argue that it's worse because now you have no idea what the actual original context was, just the AI's hallucination of it.

Re: Don't paste the AI, please

#157

Post by lelanthran »

nchmy wrote: post #76
mscbuck wrote: post #48

Earlier quoted context omitted.

Just want to say, "Copying and pasting Claude responses into Slack or an email directly shifts the burden of comprehension and understanding to everyone else" is something that I don't think I really internalized until you wrote it in that way. Which I think is evidence of what you are talking about.

[flagged]

> But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is...

I think your linked chat demonstrates quite well why you are wrong.

The sender's burden is a few lines of text "Write me $FOO, conclude with $BAR".

The recipients burden is an order of magnitude greater, trying to find your conclusion in hundreds of lines of text.

After all, your initial line:

> But isn't that "burden" on others always going to be there, whether you or AI wrote it? That's what communicating is...

Is concise and clearly understandable in 21 words . I understand your point even if I don't agree with it.

The linked chat has 1424 words. Your point is in there somewhere, but now you want the recipient to play "Where's Waldo" to find your point.

(BTW: I can't believe just how effectively you argued against your own point! I have never seen someone do it as effectively as this in all the time I have been reading HN)

Re: Don't paste the AI, please

#158

Post by croon »

nchmy wrote: post #153
croon wrote: post #139

Earlier quoted context omitted.

> The fundamental issue is whether people are taking responsibility to communicate what needs to be communicated, in an effective manner. While I agree with your overall distinction, I believe the scenario you portray is so unrealistic that you will never encounter it. Yes; if I could wrangle Claude into communicating exactly as I would have, the practical difference is of course not observable. But that's never happ…

> if I could wrangle Claude into communicating exactly as I would have, the practical difference is of course not observable. But that's never happened, and I've had an enterprise license at work for years. Sorry to say, but this is a skill issue (and not necessarily with using Ai). I've used Ai MANY times to help SIGNIFICANTLY improve things that I've written - from a few sentences to many paragraphs. As I've said i…

> Sorry to say, but this is a skill issue (and not necessarily with using Ai). I've used Ai MANY times to help SIGNIFICANTLY improve things that I've written - from a few sentences to many paragraphs. As I've said in other comments, it'll rapidly and effectively remove all my worst tendencies. And, likewise, sometimes I'll have it write a first draft when I give it appropriate context/guidance that I can't be bothered to fully curate/synthesize. Then, once it gets that draft, I'll take over and tidy it up.

Isn't this contradictory? Why do you need to take over if it's doing such a good job?

Re: Don't paste the AI, please

#160

Post by darkwater »

seer wrote: post #35

I’m not convinced tbh, a lot of my colleagues messages are “x broken” or “I do y” … no context whatsoever, and then I have to coax their context for making these decisions. Since people started using Claude now I get full context of everything… might be too much sure, but to be honest over-communicating seems better than under-communicating - sure it’s boring and tedious but that shifts the blockage to me. Otherwise…

> I’m not convinced tbh, a lot of my colleagues messages are “x broken” or “I do y” … no context whatsoever, and then I have to coax their context for making these decisions.

> Since people started using Claude now I get full context of everything… might be too much sure, but to be honest over-communicating seems better than under-communicating - sure it’s boring and tedious but that shifts the blockage to me.

Yes, but the kind of engineer that used to say "X is broken" will send you some slop with missing context and Claude will try to apply general solutions from its training data rather than from the specific context. As somebody on the Ops side that usually helps devs fixing things here and there, in those cases I'd rather them say "X is broken" and then I ask the question to Claude with all the context I have.

Probably the best solution currently is a shared skill with at least some curated context.

Post reply on HN