|
|
|
|
|
|
by Elte
1060 days ago
|
|
|
It's good to know I'm not the only one. Whenever I've had an occasion recently where I thought an LLM might be useful (either I didn't know the framework or I didn't quite understand the subject matter) and asked ChatGPT or Bard, they always just hallucinated crap that didn't work and couldn't be made to work. Just yesterday Bard hallucinated a method that was called `something_that_exists_somethingIAskedAbout()`, switching to camel case halfway through, that obviously didn't exist. For the same problem, ChatGPT came up with an existing method, but modified its parameter list to make it do things it didn't do. I struggle to think of a single instance where it's been helpful for me, and I keep trying, because I don't enjoy doing useless grunt work any more than the next guy, and I want to at least try to learn to ask it the "right" questions. |
|
|