ChatGPT made me cry and other adventures in AI land | Hacker News Reader

ChatGPT made me cry and other adventures in AI land

kottke.org

23 pointsSLHamlet17 comments

> And when I read the last line, I burst into tears. So does that make ChatGPT a soulful poet of rare ability? No. I’ve thought a lot about this and here’s what I think is going on: I was primed for an emotional response (…) and ChatGPT used the language and methods of thousands of years of writing

But that’s what’s humans do, don’t they? I’m not a poet, but if I wanted to become one, I can hardly think of way that wouldn’t involve reading a lot of stuff written by others.

Indeed. Arguments like this one immediately make me ignore everything else I just read. This is magical thinking that willfully overlooks how art works in reality. As if Shakespeare could have written Romeo and Juliet even if he had never interacted with another human, or read a single like of text written by someone else – just "from his soul". What utter nonsense.

I think the crux of the author’s statement is in the definition of “what humans do”.

You seem to be thinking that he means “read a bunch of prior work and come up with something similar” where I think that the author meant something more like “navigate the emotional landscape and write something that evokes emotions in others”.

I see several posts where people are claiming 'GPT makes me 50% more productive at coding'.

Is there any guide to how, or some livecoding stream examples? Is it just asking GPT to 'write a function which posts a url to facebook using the API' etc?

Also how to get an OpenAI login/API key from an unsupported country?

I can't share my code, but I do things like "here is my Livewire component and here is the blade view, remove these three properties and replace it with a Carbon object". And then I paste like 400 lines of code and walk around in my office. After a minute, I have sometimes 100% working code (PHP but also the frontend), that I can paste back in PHPStorm. Sometimes GPT-3.5 can do this, but sometimes I need to use GPT-4, because it "understands" much better what I want. Also, it sometimes optimizes my code, because it "saw" something, that I have not thought of. Github copilot was kind of cool, but this is magic and really helpful.

It also just writes my unit tests. Prompt is "here is my code (pasted a few hundred lines of code), write tests."

Yes I played briefly with copilot and it was amazing in some ways, and useful, but perhaps not revolutionary.

WHat you describe seems quite different. Is the chat window still the way, there must be good ide plugins by now or soon?

I think they did something with copilot x and vscode, where they have a chat like window in the IDE. But I'm using PHPstorm and since I mostly write laravel code, nothing is better than PHPstorm with the laravel idea plugin.

I also think that the chat mode is the right way to work with an AI (as long it is so slow, like it is now), so I don't mind the copy-paste workflow.

It’s trivial to write a tool that interacts with the “ChatGPT 4” API to do exactly what you’re thinking of.

The tools are coming faster than you’d think, it’s just that most of the IDEs need a couple months to go through UI design, QA, testing, debugging, documentation, and release.

I susspect there is something different at play which resulted in tears. Let me guess: deep down there, they were crying because an AI wrote that emotional poem, and not the parent themselves... But thats an insight nobody really wants to know I guess.

Kottke literally addresses that.

> When you’re really in your feelings, even the worst movie or the cheesiest song can resonate with you and move you — just the tiniest bit of narrative and sentiment can send you over the edge. ChatGPT didn’t really make me cry…I did.

Nah, that doesn't feel like they addressed what I was hinting at.

Also, strictly speaking, its always yourself that makes you cry. And touchy people, or "zu nah am Wasser gebaut" trigger that more easily, thats all.

You're right. My point was that you're wrong because he literally said what was happening.

Sure, they said they know what is going on. The essence of my statement was that they don't know yet :-) You're assuming self-reflection is always perfect, I am saying they are on a path... I don't really see a contradiction here...

Kottke, as an often self-reflective writer, is likely to know much more about his internal mental state than you do. That's my point. So it's not so much that I am assuming he has perfect self-reflection as it is that your wild speculation is much more likely to be off the mark than accurate. That's all I meant.

This is such good writing: it's thoughtful, incisive, and revealing. Before long, AI will probably be able to produce something similar, but meantime, I'm savoring that special feeling you get as a reader when a writer's words resonate deeply with you.

The most annoying things about generative pre-trained transformers is that everyone has some opinion on it, which they think is special or profound in some way.

I am sick of having to filter out internet e-celebrity number 48320's opinion on it from my front page on every single website.

Ok, but please don't snark or fulminate on HN. We're trying for something else here.

https://news.ycombinator.com/newsguidelines.html

Reply on news.ycombinator.com