With the type of AI they are creating? Probably the best you’ll get is a paper containing a ton of misinformation that sounds correct combined with already established facts/research.
GenAI can easily regurgitate anything in its training data, but you ask it something about the coding language you just created and it’ll bullshit you because it cannot work without you feeding it training data on said language. It’s very much so like a human in that you ask someone to answer a question about something outside of what they know and they cannot reliably give you a correct answer outside of sheer dumb luck.
Of course the good human will hopefully comes to not knowing.
GenAI has an unfortunate tendency to output consistent sounding stuff with high apparent confidence that means nothing at all. Which still can be useful in a context where the “correctness” can be objectively validated like certain software development tasks, but terrible for anything where the output cannot be automatically verified. Have to decide if it helps you research more concretely, the output doesn’t actually matter if it is wrong
With the type of AI they are creating? Probably the best you’ll get is a paper containing a ton of misinformation that sounds correct combined with already established facts/research.
GenAI can easily regurgitate anything in its training data, but you ask it something about the coding language you just created and it’ll bullshit you because it cannot work without you feeding it training data on said language. It’s very much so like a human in that you ask someone to answer a question about something outside of what they know and they cannot reliably give you a correct answer outside of sheer dumb luck.
Of course the good human will hopefully comes to not knowing.
GenAI has an unfortunate tendency to output consistent sounding stuff with high apparent confidence that means nothing at all. Which still can be useful in a context where the “correctness” can be objectively validated like certain software development tasks, but terrible for anything where the output cannot be automatically verified. Have to decide if it helps you research more concretely, the output doesn’t actually matter if it is wrong