Responsibly empowering developers with AI on MDN
blog.mozilla.orgThey seemed to have missed a very important piece of critical feedback in the GitHub issue: absolute beginner users don't know enough to know if the AI is being helpful or not.
They'll click the thumbs up button and then realize hours to days later that they were taken down the wrong path. Or they'll just have a general feeling that everything is frustrating and CSS (or whatever) just don't make sense or work.
But they'll never be the ones saying that the AI mixed up grid cells and grid rows. That'll always be more experienced people, who Mozilla then dismisses as people looking to trip up the AI rather than people genuinely asking.
For certain definitions of responsible. I don’t think this class of model can be deployed responsibly in this context, because it has no guarantees on correctness.
Likely related:
Another related issue:
The AI help button is very good but it links to a feature that should not exist