2 ms·

> I'm puzzled why anyone would incorporate code on trust. Because almost every programmer these days has learned by a route that relies on incorporating code o

> I'm puzzled why anyone would incorporate code on trust. Because almost every programmer these days has learned by a route that relies on incorporating code on trust. E.g. using someone else's compiler. > I use code from LLMs on most work days ... not because I trust it, but because I understand it and tested it. The snag there is that one can rarely verify test coverage is sufficient. > the same as code from anyone, including from myself six months ago. But it is not the same. ChatGPT delivers completely untested code. Not even syntax-checked in my case. Would you accept such code from any human? > Trust doesn't enter into it. Clear and workable solutions do Plenty of clear and workable solutions are discovered unsafe. I suspect that's much more likely when the code comes from a source such as ChatGPT that delivers code in a completely untested state.

Brian_K_White 3y ago
Syntactically incorrect code is not much different than an outline or pseudocode. There are way too countless many stories of people successfully accomplishing tasks by getting an llm to give them a starting point or outline or even a complete solution that only needed a few fixups, hardly any different from debugging your own first draft, to say it doesn't or can't work. That's already sailed. The fact that it doesn't produce finished correct code, or even appropriate outline for every single case, doesn't seem especially remarkable or damning to me. It is what it is, it's not what it's not. (I haven't used any myself, but I'm not primarily a coder who is obligated to take advantage of every available tool, and don't like how any of the llm companies are training on open source code without either consent or attribution and then selling that. And don't like how ai and even mere speech to text voice interface before that is being added to everything in general.)
- dllthomas 3y ago
  > hardly any different from debugging your own first draft I find this take surprising. Leaving AI aside entirely, debugging my first draft seems very different than debugging your first draft.
  - Brian_K_White 3y ago
    I often don't know what's wrong with my broken code. I only know what I want, not exactly in what way I failed to produce it. I have to debug to figure that out. I grant, maybe that is still quite different. There is something extra to grasp the theory behind someone else's code sometimes.

Brian_K_White 3y ago

Syntactically incorrect code is not much different than an outline or pseudocode. There are way too countless many stories of people successfully accomplishing tasks by getting an llm to give them a starting point or outline or even a complete solution that only needed a few fixups, hardly any different from debugging your own first draft, to say it doesn't or can't work. That's already sailed. The fact that it doesn't produce finished correct code, or even appropriate outline for every single case, doesn't seem especially remarkable or damning to me. It is what it is, it's not what it's not. (I haven't used any myself, but I'm not primarily a coder who is obligated to take advantage of every available tool, and don't like how any of the llm companies are training on open source code without either consent or attribution and then selling that. And don't like how ai and even mere speech to text voice interface before that is being added to everything in general.)

dllthomas 3y ago
> hardly any different from debugging your own first draft I find this take surprising. Leaving AI aside entirely, debugging my first draft seems very different than debugging your first draft.
- Brian_K_White 3y ago
  I often don't know what's wrong with my broken code. I only know what I want, not exactly in what way I failed to produce it. I have to debug to figure that out. I grant, maybe that is still quite different. There is something extra to grasp the theory behind someone else's code sometimes.