You are viewing a single comment's thread from:

RE: New Trending Algorithm - Attempt Number 5

in Steem Dev6 days ago

I usually ignore the page too :-)

but it feels even more abrasive than the default trending page does when valueless posts make their way into the list.

Does that mean you still don't like it overall?

The key question is surely how we would define ‘trending’. I think you're more likely to see Thoth as a way of identifying valuable posts. Unfortunately, that's not readily feasible with Hivemind...
But now that I think about it, it seems theoretically feasible in Hivemind to have an AI evaluate the posts and interactions... if there is an appropriate interface and resources for this. Of course, it has to happen very quickly so that the blocks can still be processed promptly...

Since several users (including myself) have already written that they don't use the site at all, I wonder whether it's worth the effort at all. Perhaps simply changing the start page will help. :-)

Sort:  
 5 days ago 

Does that mean you still don't like it overall?

No, it definitely doesn't mean that.

I think this is an important change because it gives people a way to find new authors to follow. Also, I just noticed, today, that there was something funny going on with my browser, so I wasn't seeing the right set of posts. Somehow, it was set to use "api2.steemnet.org" instead of "api.steemnet.org". Now that I switched, it looks much better.

I think you're more likely to see Thoth as a way of identifying valuable posts. Unfortunately, that's not readily feasible with Hivemind...

Yeah, I don't think an LLM solution would be suited for the trending page - because of the speed issue that you mention (and also cost). Using machine learning to adjust weights based on webserver log data might be a good use of AI, but obviously that work effort is a different order of magnitude. Not suited for a first effort, I don't think.

Since several users (including myself) have already written that they don't use the site at all, I wonder whether it's worth the effort at all. Perhaps simply changing the start page will help. :-)

I think it's worth the effort.

  1. As already mentioned, it gives us a way to find accounts that are worth following.
  2. With a hivemind implementation, different site operators can compete for users by offering the best trending algorithm. This would encourage more decentralization. (although that might complicate things with the public APIs and loadbalancers - developers who make use of the /trending feeds would need to pay careful attention to their api choices)
  3. The communities trending is also improved, and I use that a lot.
 20 hours ago 

Thank you.
I've had a lot on my plate in the last few days and unfortunately I haven't really been able to get to grips with the problem. But I really appreciate your answer.

!dubby 20%

Thank you for sharing quality content on Steemit
You have been supported by the team:


Curated by: @kafio

Thank you for sharing quality content on Steemit
You have been supported by the team:


Curated by: @kafio