[ivanvoid]

I genuinely don’t understand how anyone (with technical background) can see LLMs anything more then fancy autocomplete. If you know anything about NNs and about average code quality, that LLMs never will be able to generate high quality code.

Im ready to get downvotes again for my takes, but as a person who writes and trains DL models, I will die on the hill: “people need to produce high quality data” it can be code it can be art, but we can’t rely on those models and trust in the things that they provide.

[s_dev]

The new bottle neck isn't writing code, it's testing. You're right you can't blindly trust the output of an LLM but you can trust the testing regime to ensure a certain standard has been met. In hindsight this actually sort of obvious, the more things change the more they stay the same etc.

[ivanvoid]

Well it’s not obvious that is true. If you ask LLM to write tests, it will generate versions of them that code passes, that doesn’t guarantee good code. If you write tests yourself and just pray for great LLM pull, it’s easier to just write code yourself, in my humble opinion

[s_dev]

That's a useless approach as you point out but doesn't meant there isn't a valid testing regime to be explored and upheld. Manual testing is going to be a lot more important, I see QA teams/roles becoming very valuable assets in the coming years.

[em-bee]

you can trust the tests that you have written, but what about the tests that you didn't write? can you be sure that your testsuite is complete?

when i do test driven development, all the thinking goes into the tests, and the actual code writes itself. LLMs hardly help make that any faster.

having a complete testsuite may make it easier to use LLMs for refactoring, and adding features, but then you still have to write he tests for the new functionality.

[roncesvalles]

Because often the problems that these people are working on tend to be trivial. LLMs are excellent for making the millionth CRUD backend server that talks to SQLite and a glorified todo list React frontend. In fact, it's stupid to do it any other way now.

[WithinReason]

> If you know anything about NNs and about average code quality, that LLMs never will be able to generate high quality code.

I recommend looking into a subject called "reinforcement learning", the way AI acquired superhuman skills in chess, go, etc.

[ivanvoid]

Obviously I familiar with RL, written multiple training pipelines in my day. and in order to gain that “super human skill” using RL you need to define fit functions and provide environments that will provide you with feedback that used for training. Go and chess are have clear rules and environment that provide you with a signal of success, I waiting to see this for coding, I don’t say it’s impossible just orders of magnitude harder