Which it does
[–] Razengan 2 days ago
[–] g-b-r 2 days ago
?
> Not worse than humans have been doing on their own since forever
If it's not reviewed well (some people stopped even looking at the code) it can be immensely worse.
And most people end up producing much more code with LLMs, quickly reducing how much of it they review; a lot have become fine with turning the interface upside down often just because it's easy, even if the new interface is worse and a lot buggier.
[–] Razengan 2 days ago
When was the last time you ..or anyone.. reviewed the assembly language or binary output of your compiler?
...
What if the CODE DOESN'T MATTER anymore?
No really, wait:
What if all that matters is the result?
First, you ask the AI to make what looks correct.
Then you ask it to make it work correctly.
Then you ask it (or another adversarial AI) to do a security review pass.
Then a performance benchmark.
Then a few more iterations of whatever.
If all pass, does any human really NEED to look at the code?
[–] rafaelRiv 2 days ago
not even a week ago. But generally people prefer to burn millions on compute instead of seeing the output of the compiler and see how they can optimise their code
[–] g-b-r 2 days ago
An LLM can spout anything at times, however good it might be on average.
> If all pass, does any human really NEED to look at the code?
Yes, for the love of god.
You don't really know what's the result: however comprehensive and correct your test suite might be, however well you check the running software, there are things that are unlikely to be caught by that, such as data sent to unintended servers or deletions of random files (maybe only under weird conditions).
Having other AIs review the code is not going to always catch everything, and pleading the AI to do a good job won't help very much.
[–] Razengan 2 days ago
How many bugs and vulnerabilities from unterminated strings and buffer overruns and underruns? How many leaked accounts and passwords?
What's the longest you've gone without hearing about a 100%-human-caused bug? Why wasn't it caught? Why did it happen in the first place?
[–] g-b-r 2 days ago
The point is that a (well-meaning) human will never add obviously malicious code, while an LLM most certainly can do that.
> Which it does
Not worse than humans have been doing on their own since forever.
I've been using AI to make private apps for my own personal use, and it's pretty amazing how quickly I can iterate to fix minor niggles and UI annoyances.. within minutes, while the app is running.
Meanwhile operating systems, YouTube, Facebook, Reddit, etc have such shit UX here and there for YEARS. You're telling me they can't fix it with all their employees that get paid hundreds of thousands of dollars?