On Bots

in #curation8 years ago (edited)


[Image Source: pixabay.com, Licence: CC0, Public Domain]

This will probably be my last post on the topic ( unless I think of something else to say ; -), and it's only for my own benefit, but I want to save all these links in one place so I don't need to scroll back through my post/resteem history when I need to refer to one of them in the comments. There are a number of ideas that come up repeatedly when the question of bots arises.

In short, my position continues to be that in the long run, a combination of human and automated curators will do a better job at discovering quality posts than humans alone. Bots are just tools, and a tool can be used well or used badly. It is a mistake to blame bots for bad decisions that are made by some human operators. It is also a mistake to think that what we see from bots after just a few months of development is what we will always see. Here are the things I've written on the topic:


Promoting from Comment, Feb 14, 2017

It's important to remember that we're looking at first generation bots. In the long run, the bots that succeed will be the ones that promote steem's long term value. A short list of things bots can do that human curators can't/won't:

  • Judge all articles using a single, consistent standard.
  • Mine statistical correlations between post content/metadata and steem value
  • Work 24x7x365
  • Level the playing field for authors between short posts which humans view quickly (10 memes per minute per human?) and longer posts which are time consuming to read manually (some longer posts can take 5-15 minutes to read carefully).
  • Check for plagiarism
  • Check for repeated posts
  • Evaluate posts in multiple languages.

Of course there's more, but hopefully that gets the point across.

Sort:  

I see it in a very similar way. It is not "them" or "us", it is "them + us"
Because 1+1 is way more than 2 :-)

Exactly. Humans and bots have complementary skills that create a positive feedback loop. Input from humans can help bots to make better decisions, and voting by bots can help humans to decide where to focus our valuable attention.

In a tech-savvy community, it amazes me that this is controversial.

a combination of human and automated curators


Curate: select, organize, and present (online content, merchandise, information, etc.), typically using professional or expert knowledge.
"people not only want to connect when using a network but they also enjoy getting credit for sharing or curating information."

You need a kind of AI not even google has for that. Not to mention the kiddiescripters that overpopulate steemit thanks to piston's great functionality.

Call them "voters," if you'd prefer. Then curation is the end result of combining inputs from all the different voters. I certainly don't have professional or expert knowledge about every post that I vote on manually. I doubt if many people do.

A bot is merely an approximation of the operator's own preferences. The closer the approximation, the better the bot. So yes, there is plenty of room for improvement in what we see today, but curation rewards create an incentive for continuous improvement. I'm confident that it will happen.

So far, the only measuring method I've seen from bots is reading the payouts of previous posts to check if the curation reward will be high enough to be worthy of an upvote.

Flipping a coin is not curating.
Following the trends of other bots is not curating.

You are right, it should not be called curation.

PS: "typically using professional or expert knowledge." =! "always". You only need a certain level of interest to curate here, and a bit of criteria; both things totally absent in votebots.

Yes, bots are tools... But, for now people is using screwdrivers to nail boards.

My own bot doesn't even look at payouts. I check its trail regularly, and if I find it voting on something I don't like, I adjust the rules accordingly. Over time, flipping a coin kind-of is curating, because you're doing something like a random walk with error correction to find progressively better approximations for your preferences.

There's no correction in the bots I'm referring to. Just a plain series of IFs, or even worse, a list of "to dos" (like steemvoter.com's).

Your bot, would be the 1st one I see with a small NN that actually works.

I wouldn't go so far as to say it works. Just that I'm always trying to improve it. ; -) I believe that steemvoter plans to introduce premium features that will improve the quality of their clients' voting, too.