I am a huge fan of nearly everything HF does. I am however slightly dissatisfiedd by Spaces of the week now…
The change from showing most recent update (which shows me new spaces people are jamming with) and the new Trending ordering as default which lists ‘popular’ AI spaces, is one area I see as having potentially side effects of not properly encouraging new development or exposure of AI spaces that might be usable and performant.
Instead I feel it is favoring larger continual investment AI’s based on likes. Maybe spaces with GPU which is okay - I like those as long as they work and prove it is state of the art and completes.
Potentially currently the new ordering by likes is exposing lots of non functional AI’s which are quite frankly are duds when they repeatedly fail to yield any output.
The fact that lots of people may have submitted their like vote, in my opinion is not as good as the most recently updated or potentially high check in to git spaces which I want to see!
A new algorithm is needed like performance in time which makes sense to recognize or discover the most innovative new spaces which are performant and return results but are being suppressed by over focus on popular, much like our problems in social media where advertising and capture exclude the indie developers making great strides (this I care about seeing more than anything) and lead to people abanddoning pllatforms that dont have high return on effort investment with some form of recognition.
For example, last week I checked all eight spaces on Trending list and found like 6 out of 8 failing on test by a newcomer despite that they all had lots of likes. The two which worked were amazing of course so it is definitely not completely broken.
Popularity contests are not good for innovation in my opinion however from a model perspective space download count makes sense.
I feel (entirely my opinion) that we were better served by spaces that were sorted based on a combination of factors for instance:
-
Spaces where if you submit new inputs do not give any results or error within 2 minutes of submit should not be on trending spaces, even if they had lots of likes. Why? They do not address pains of users, don’t express joys and only reflect like count.
-
I would rather see: Spaces that have repeated check in count. This to me is obvious to see that codde actively being improved is our collective state of art, not always just what has been historically.
-
The fact that an unmaintained space dies is sad to me. HF also if we could have some form of ‘rebuild all’ would be great to reignite work if it had at least one update per month in recent times rather than automatically stopping it… I have had a few favorites of others that died out this way.
HF has been a bastion of freedom and equity in AI when it sheds light on 1. What works now. 2. What changes the game. and 3. How immediately useful it could be to a user that needed just that.
Please reconsider how to structure spaces in Hubs that have high turn around daily on edits and show those or give an alternative view for us that want to discover the latest and greatest.
I feel that the latest and greatest sometimes still show amazing work and I want to see those but lately with current algorithm I am only finding maybe 1-3 out of eight that actually complete AI loop after 2 minutes. Frankly if no turn around is given to users that is not a useful space.
For comparison and judging here are four spaces I worked on this weekend. Previously I would have seen these repeatedly on last modified list (they way it was before) which if getting a follow up feedback in community would keep a developer working hard at making it better just under knowledge of it being used.
I also see that we need to support the spaces where not only does it give a great result within reasonable time every time (say 1 minute is the limbo stick) it also continues to be the best we have in a particular task. Because of this I feel you are currently overlooking lots of task specific AI that is continually improving yet gets neglected since its not in top eight.
Thanks too for being my favorite AI platform by far. You all are amazing.
Much love,
Aaron