Notices by Antoine Leblanc (nicuveo@tech.lgbt)

@r3trofitted thank you, i appreciate it! i wasn't sure at first whether it was right for me to say anything, since i'm not a heavy user of the MDN in the first place... but the more i look into it, the more i am convinced this "feature" reveals deep organizational issues.

i think they suffer as well from the "exposure" of being one of the few bad "AI" projects that is being debated in a public forum. if Google decides to train Bard on all my docs, and insists on putting generative stuff everywhere in GSuite, there's little avenue for me to join others in criticizing that decision... but here it's all on GitHub, and anyone can jump in.

and while i worry that it is becoming a bit of a pile-on, which is why i am not planning to engage further, it is (i think) very revealing that almost no one has jumped into the fray to defend the MDN so far...

oh my god i had missed this from the postmortem, but the way they validated the test queries they ran (asking the LLM for explanations of code samples) was... TO FEED THE RESULTS INTO GPT-4 AND ASK IT TO CLASSIFY THEM AS "ACCURATE" OR "INACCURATE". THEY ONLY REVIEWED SOME OF THE ONES LABELLED AS "INACCURATE".

i want to scream.

at least, i feel vindicated for engaging in the thread despite not being a heavy MDN user, because it is extremely apparent that they have no idea what they're doing.

the answers (or rather the lack thereof) in the ongoing MDN issue (https://github.com/mdn/yari/issues/9230) continue to be a source of headaches. it is transparently obvious that the reason they want to introduce LLMs is not for the users' benefit, but for some other incentive they seem unwilling to disclose...

but also the sentence "incorrect information can still be useful" has been uttered, and i have yet to recover from that.