RE: Steembay (a bot) under attack by a wannabe AI

You are viewing a single comment's thread from:

Steembay (a bot) under attack by a wannabe AI

in flags •  6 years ago 

I know what you mean, but as I said in another comment we don't have to do everything, this is just one thing and I think the rationale is defensible. There is relatively low SP behind sadkitten accounts and friends currently.

For example @haejin did appear on our list last week in 51st place. When we improve the algorithm to not only take into consideration self voting but clique voting we will see the focus change a bit and I expect the bot will be more interested in even more problematic cases.

The bottom line is that there's nothing controversial going on here, @steembay are massively self voting. Even a little less self voting and the bot doesn't care about them.

Authors get paid when people like you upvote their post.
If you enjoyed what you read here, create your account today and start earning FREE STEEM!
Sort Order:  

@personz I hadn't really looked into @sadkitten till today but I like the concept. It would be great to see this bot growing to tackle some of the bigger fish who are probably getting around the self voting in other ways. I am sure you will get there.

I was curious as to the feedback your getting on the bot and if your tracking how it is changing peoples voting patterns. Has anyone done some analysis on this?

It would be great to see this bot growing to tackle some of the bigger fish who are probably getting around the self voting in other ways.

We're currently on v3 of the algorithm, and all versions so far have focused only on the self votes of the user. In this version we were able to pretty much perfect a behavioral model instead of a value based model, so with that working well the next step moving to v4 is to take all account votes into account. Self voting will no longer be the strict focus then, but variance in votees and behavior around that. So if an account only votes for one other account, it will be rightfully recognized as effectively the same as self voting. And that will scale up.

I get a lot of positive feedback in private, and we've always had some opposition from people who disagree with it, such is the kind of venture it is. It's hard to know but several large stakeholders, developers and people I respect and think highly of are backers, which encourages me.

I did attempt to reach out @paulag recently to ask her to do an analysis. I don't have the time right now to do one except my own internal metrics. I was intending to publish something soon and @transisto and I are still discussing the best way to do this.

What is it you'd like to see in particular?

This post is one big data point, where a person feels that the bot is taking away his right to rewards. It would be interesting to see if other users felt the same and wrote a post, or if they adjusted their voting behaviours as a result of the bot downvote.

It would be also good to track repeat offenders and and what level of SP Holding it took for people to take notice of the downvote. The whales surely don't care at the moment as the downvote is inconsequential to them.

If we could get out of the analysis what cohort of people are affected by this bot and what cohort take notice of it that would be very interesting.

I cant wait to V4 as I think that will greatly improve the effectiveness of the bot.

Me too, I'm working on it now, moving slowly and carefully to make sure we get it right. There will be a lot of testing.

This week starting today marked the start of the 6th week where I have good records of all the data. I will do some kind of report soon, I think I will aim to do one at the end of week six, so in one week. I could also summarize some of the comments we've been getting, and the type of arguments people make against us and our concise response to those.

We will also be launching some kind of method for people to contribute their accounts to the bot when their VP is "idling" at 98+%, but that also needs to be done carefully, of course.

Thanks for your interest and spurring me on to get out some kind of report! Some others have been asking too.

Oh? Clique Voting!?

What percentage of random votes do we need to stay out of the badkitten crosshair?