Hacker News Reader

Comment by ryanklee | Hacker News Reader

What matters is whether the suggestions are actually good, not where they came from.

The author makes it clear in the associated discussion that he doesn't immediately assume the suggestions are bad based on their source:

> I take the time to understand and consider each suggestion, not rejecting anything out of hand, and share them with my team members (of which there are 17).

The issue is the time it takes to explain _why_ these are bad ideas to non-technical (and skeptical) colleagues.

That sounds to me less of an issue with ChatGPT, and more with having colleagues that don't understand how to engage in reasonable discussion or evaluate information correctly.

Good point. Maybe they can invite a discussion where the suggester walks through ChatGPT steps explaining how it is done? Hopefully people aren't so flippant they hammer their keyboard and hit send email. If that's the case, then maybe they need an ego check.

I think my default response would be (regardless of the ChatGPT instructions):

"I looked it over and it doesn't actually work. ChatGPT is good for many high level things but when the specifics get technical, is struggles and invents solutions that don't exist; like a bloom filter. Ask ChaGPT what a bloom filter is."

And the question was how to get the their coworkers to stop giving copy/pasted ChatGPT technical answers from prompts written by nontechnical people, not how ChatGPT can do a better job.

And I didn't suggest how to do so. I implied that there is an issue with the quality of their coworkers, not ChatGPT.

And since nobody implied it was a problem with ChatGPT itself, I pointed that out. Glad we cleared all of this up.

The saltines is basically inappropriate.

Apologies of it read as salty rather than mildly tongue-in-cheek.

ChatGPT is absolutely a symptom here of the underlying problem. Frankly I think it's possible that OP is being too solicitous with these requests. Why does this other team think OP needs a constant input of new ideas? Why do they feel comfortable repeatedly second guessing OP in OP's are of expertise? It sounds like the planning process is totally out of whack. If they were clipping ideas out of a magazine and asking OP to do them it could be just as bad.

It is significantly cheaper to ban a poor tool than to teach critical thinking.

The problem is with the person, not the tool. Get rid of people who can't think critically, replace them with people who can.

Sure it's not the tool if you completely ignore 80% of the story. It's cheaper to ban a tool than "hire people who can critically think", as if I ever believed people could intentionally detect that.

brandolini's law

intellectual entropy

The issue here is that the artists are confident that the suggestions (which they don't themselves understand) are good because they came from ChatGPT.

I would argue that if you cannot answer questions on something generated before suggesting it or handing it off, you absolutely should not be pushing for it. Do the research necessary to feel fully confident on a hand-off.

Exactly!

GPT can be a fantastic time saver, but it is unable to check itself.

You absolutely need to do the research to check it and correct its output.

It's like a power tool: potentially very powerful, but it'll make a huge mess if you let it run out of control.

Hell, if something doesn't make sense, ask it again for clarity. It does a reasonable job at that, but again, before you hand off a solution, make darn sure you understand it, because chances are high, whoever pitches a solution that sounds usable, is going to break ground and work it.

Honestly had the same experience with a junior developer yolo-ing some changes into a PR before even running it themselves to check that it works. It's just a waste of everyone's time and a lack of respect. That's what needs to be conveyed to the artist colleagues.

They aren't good. Also, diffusion models work well for the artists to spit out pixels. The artists assume the LLM generated code is the same quality and that the OP is a fool who won't do what they ask due to lack of skill or stubbornness.

It's messing up the dynamic where creatives come up with blue sky stuff and developers come to a compromise on a possible solution. Now you have this AI model hallucinating plausible, but fake solutions.

The model says what they want to hear because it is a chicken, not a pig in this scenario.

the issue is it takes effort to determine if the idea is good or finding subtle errors in generated code, generating it with GPT requires almost no effort from the person who then offloads it

Basically, a non-malicious https://en.wikipedia.org/wiki/Gish_gallop

My gut reaction is no, but let me think out loud using ad absurdum to see if there is merit.

A five year old asks chatGPT how to achieve world peace. The response is not only possible but easily affordable on a short timeline. Do I care that a 5 year old got it from ChatGPT? I guess not. A part of me would want all the adults on the planet to stop and acknowledge how tragically ineffectual they are, but as far as whether the source matters...nope.

Good point. Thanks for letting me think through it.

Where they came from is part of how you decide if they are good. Especially for nuanced knowledge work.

If not, I have a pacemaker I would like to sell you.

Yeah, but it appears this guy can't convince the suggestion givers because they lack the expertise to evaluate their suggestions. For instance, they could be asking ChatGPT how to do something in code and then sending that.

Reply on news.ycombinator.com