Recently i’ve lost a lot of interest in submitting title or description edits because so many are rejected by ML. I could never keep up by using appeals to correct that mistake (even if i wanted to use appeals for that) so it just isn’t worth my time.
If you’re encouraging us to submit a lot of edits then please make it worthwhile and at least give any edit that ML doesn’t approve the chance to get some human eyeballs on it.
Thanks!
Being there’s community rewards for edits approved by ML, i doubt this will happen.
ML accepting is fine, any it would already accept it should accept. but instead of rejecting, it should push it to community review
I feel like my post was pretty clear about the request.
This whole exercise is to test the AI, and let it do ML.
Niantic said “pretty please”, so I’ll help. Unless it gets too impossible…
ok. that also doesn’t have any bearing on what I asked for.
flag the ones that it would have rejected and review them but treat them the same as the ones it makes no decision on currently.
anyone else want to post about something that’s not what I posted about?
Agreed, if eMiLy can’t see what the edit is, eMiLy should not reject. It should just get moved to community voting. Then it should be able to help eMiLy learn what good edits are.
I’m hoping that is what is happening.
Highly doubt
Have you tried appealing?
He said he can’t keep up on the appeals, I’ve run into this as well. Something that locals should know or be able to see why I made the edit, gets rejected by eMiLy when if it went to the community they probably would know more. Hard to appeal it when we don’t have many actually easier to just edit again and hope for a different outcome.
Imagine thinking…
So far 4 out of 5 were accepted in the first day for me.
I have only had 3 text edits accepted, 2 being title edits fixing grammar. I also submitted a few descriptions at Wayspots without any, and none of those were accepted by ML, even though they were quite detailed.
All my photos though, ML has accepted. Go figure…
Yes. AI gives an answer (pass/fail) and a confidence level (%). I think Niantic has dialed up the auto-apply confidence level, from around 80% confident to around 95%. That’s what the manuals say to do in Machine Learning phases. Verify those less than 95% (or sumtin) confidence level. Once the wave testing is done, set it back to something like 80%.
The AI seems a bit more lenient for the event, at least. It hasn’t rejected my attempts to add info to some portals yet, and it used to do that to me before the event.
Actually, here’s an idea: Why not have an AI rejection megathread where we can post examples of AI rejections that we feel should’ve been community votes instead?
