[RamblingCTO]

I'm not scared. ChatGPT produces shitty and insecure code which nobody should just trust.

/edit: also, most of the stuff I do is so fringe that chatGPT probably doesn't know about. I'm currently upgrading our spring boot/security stuff to 3/6 and ACL is broken. There is no single sample or SO question on the internet for that. So how can chatGPT solve this? Answer: it won't.

[Xenoamorphous]

> ChatGPT produces shitty and insecure code which nobody should just trust.

Today. What about a couple of years from now? We're seeing stuff that was sci-fi 5 years ago, who knows what we'll have in another 5 years? I think that pretty much everyone that's not very close to retirement and whose job is performed while sitting on a chair should be concerned about a not so distant future where their job either disappears or is transformed radically.

[layer8]

Extrapolating the future from the past doesn’t work like that. Otherwise we’d have AR glasses, actual self-driving cars, and use fusion power by now.

[Xenoamorphous]

That's why I used the expression "who knows what we'll have in another 5 years". It's the possibility of AI keeping advancing at this pace what scares me.

[RamblingCTO]

I'm not so sure. I've been studying that stuff for a decade now and we've seen multiple of these "AI gold rushes" where afterwards an "AI winter" followed. Let's see and find out.

[RamblingCTO]

I don't think we'll see knowledge transfer or intrapolation from non-existent data, whhich is what would not scare but excite me. Current machine learning just extrapolates from the data it has seen. Garbage in, garbage out. No data in, randomness out. So I'm not scared

[qumpis]

How do you define extrapolation at the scale LLMs operate? Even if you work with unseen-to-model software, it seems sufficient to "understand" the documentation and code examples to orient itself to helpful context.

That's why I'm scared. Once embedding whole codebases becomes a viability, I expect many opinions to change too

[RamblingCTO]

I invite you to play around with the code it generates. For example, yesterday I tasked it to generate a gpt-3.5-turbo API client. In the time I needed to get it running I could've wrote it myself. And that's < 100 lines. Don't even get me started on architecture, contextual decisions, clean code etc.

[qumpis]

Interesting that it's so insufficient for you, maybe indeed stuff you do is novel and would require lots of instructing before helping you.

Personally, my usecases involve quite standalone applications. copy pasted from another thread (using gpt4):

My personal use of gpt4 (also daily) is: correct, rephrase spelling from my brain dump, make python plots (stylize, convert, add subplots, labels, handle indexing when things get inverted), make short shell scripts (generated 2FA, login vpn through console using 2fa, make script of disabling keyboard etc), and help debug my code (my situation is this, here's some code, what do you suggest?).

[RamblingCTO]

I would agree with summarization and NLP-driven tasks and I will actually add chatGPT to my side-gig for that. But code-wise it's not as big of a help as I'd like.

Good for you that it actually helps with coding! I like copilot a lot for auto-completion tho, but the rest of my work apparently is too complex, yes.

[qumpis]

I felt similarly before I started using GPT 4. Then I got scared

[RamblingCTO]

I wanna play around with "baby/teenageAGI" today, but I don't have high hopes. chatGPT hallucinates stuff together, nothing more. Would be cool if it could solve small jira tickets but I don't think it will be helpful.

[cleanchit]

> chatGPT hallucinates stuff together

If you had a creator, he would probably say the same thing about you.

[RamblingCTO]

Yeah, but I can understand unseen data, whereas machine learning can't afaik.

[circuit10]

It clearly can, it’s not like ChatGPT just refuses to answer if your prompt isn’t exactly word for word the same as something in its training data. Otherwise it would just be a search engine

[RamblingCTO]

"Can" is a bit of a stretch imho. It can produce something for every input. Humans are way more accurate in that regard. ChatGPT just feels correct but is mostly wrong.