csdummi

csdummi, 9 days ago

@molly0xfff do you ever follow-up on some of the stories you post on @web3isgreat ?

A theft of million(s) of dollars would bankrupt most buisnesses and I'd like to understand what the typical aftermath, the consequences of these grifts/thefts/scams are for the projects?

csdummi, 14 days ago

@molly0xfff I know this is completely hypothetical and the damage caused by Binance is (still) far less than FTX, but: would this have been a deal that SBF could have struck or something in it's ball park, if he'd cooperated from the start?

csdummi, 15 days ago

@Wahlen_DE warum.ist LINKE neu?

csdummi, 18 days ago

@jwildeboer @EU_Commission @EDPS there was talk of this at the #37c3 in December. It seems the problem then was: there is nobody in the EU agencies willing take over running the instances & the EDPS can't run these servers perpetually as that'd require them to regulate themselves.

csdummi, 19 days ago

@8124 Nasrallah?

csdummi, 1 month ago

8 boost, 11 favorites. Not a single reply.

Either this post and the article referenced was perfect or people just aren't engaging with the contents of the posts here like they used to.

When I posted this article originally, there was a lively discussion here that led me to summarize some of the very good and interesting points made back then in a follow-up post.

Now, in the wake of a disaster caused by the problem this article is addressing, we don't have the time on this network anymore to engage with ideas beyond a mere press of a button?

I'm not accussing anyone who boosted this or favorited this individually - if I was I'd have to find myself guilty as well. It is just a sorry development to me - perhaps there is just too much "content" on here by now for anyone to meaningfully and critically engage with any single piece of it.

csdummi, 1 month ago

@dansup I've not used TikTok or YT Shorts, so I don't quite understand what is meant by an algorithm tailored to personality of the consumer.

How does an algorithm exhibit this behavior?

csdummi, 1 month ago

@3dcandy @dansup

That's been the mode of Mastodon now for
years, and while there are good arguments for such a strictly chronological "following" feed algorithm, I think it'd be valuable to experiment with alternative feed algorithms, especially in an open source project.

Otherwise the project is quite opinionated on what the right feed algorithm is. I think we can experiment with other suggestion algorithms and find some potentially quite interesting algorithm options in the process.

csdummi, 1 month ago

@dansup got it. Those are the inputs, the big question of course is how then it reaches a decision on what the next post will be.

Could these data points be used to derive a certain "interest" level in different accounts, topics, content (in the later two cases, how'd they be identified?) and then find more or less of that account/topic/content type to show next?

A "good" algorithm in this sense'd probably have to classify the content in it's database a whole lot more, to be able to suggest "similar" or "different" content.

csdummi, 1 month ago (edited 1 month ago)

@shekinahcancook @3dcandy @dansup

I understand the concern and ads, tracking, etc would indeed be steps towards twitterifying Fedi. I don't count feed algorithms among these.

To me strict chronological feed is just as much an algorithm as any other and there are many other possible algorithms. And the Fediverse hasn't really experimented with this choice as much as with other aspects of the experience here, exactly not to "twitterify" us.

I think this is a disproportionate fear that prevents all experimenting with anything but the default choice. (And who says that is the default?)

The chronological feed is a choice, an algorithm, and to change this algorithm to something else does not make us automatically into Twitter. I'm not proposing to implement the twitter algorithm or any other proprietory algorithm in a Fediverse. But more diversity in how we choose to sort our feeds is an unused potential of the fedi and ought to be explored more.

csdummi, 1 month ago

@shekinahcancook @3dcandy @dansup

I don't think we are that far apart on this point. I agree with you that any algorithm should be under the control of the people seeing and creating contents.

My contention is that we can enable more control for these people through more advanced analysis of the content.

Presently you can only curate on Mastodon based on self-declared attributes by the author - hashtags, accounts, etc. But if some more advanced analysis and classification was performed, you might in the future also be able to curate your feed based on classifications that the author does not make explicit. E.g. subscribing to a certain topic, language, news story or blocking hate speech, explicit content or news stories you don't have the mental capacity to deal with at the moment (NSFW, Hate Speech, Antisemitism, etc. etc.).

I'm not saying an algorithm should hide something you explicity want to see, but that we could have more complex ways of saying what we want to see and what not.

csdummi, 1 month ago (edited 1 month ago)

@shekinahcancook @3dcandy @dansup and that might be the crucial thing:

As an author, I maybe opposed to analysis of my content taking place. But as a viewer I absolutely want that the content I see has been analyzed before - and this is in fact already happening.

It is just (mostly) human labor producing this analysis. The moderators of #babka and any other moderated Fediverse instance are constantly working to analyze the contents of posts from all across the Fediverse and decide whether you or me, as members of #babka, should see them. And in far too many cases they have to decide that some content must not appear on the timelines of #Babka members, delete the post here and perhaps block or silence the author or even author's instance.

This is content analysis and it is crucial for this instance to exist.

csdummi, 1 month ago

@shekinahcancook @3dcandy @dansup I'm not enamored by technology. I know what toll moderation work takes on the moderators here. And to deny us the use of technology to deal with waves of antisemitism, spam, racism, hate speech, is putting this instance at a disadvantage compared with the antisemites, spammers, racists, etc. - who are absolutely not producing their hate only with human labour.

csdummi, 1 month ago

@osma @dansup that makes sense - content analysis is harder than metadata analysis - but I fear that this type of metadata won't be available in a federated system (at least none of the AP implementations I know divulge watched data to other servers) and so, content analysis might be the only way to do this here.

csdummi, 1 month ago

@shekinahcancook @bradley these categories do not work. My father is a boomer and I'd be classified as GenZ. These categories are so general that they allow almost nothing to be said about any of them without qualification or specification.

csdummi, 2 months ago

@MastodonEngineering and why exactly are these being released unannounced and on a Friday?

csdummi, 3 months ago

@serge many of the projects we use in Python, like #Flask, are documented using #Sphinx, which can be configured to automatically generate API documentation from Python source code:

https://www.sphinx-doc.org/en/master/tutorial/automatic-doc-generation.html

This is (similar to the #yardoc and #rdoc situation) backwards compatible with #pydoc, though much more expressive due to the machine-readable format for defining parameters, return values and potential exceptions:

https://www.sphinx-doc.org/en/master/tutorial/automatic-doc-generation.html

https://docs.python.org/3/library/pydoc.html

csdummi, 3 months ago

Thanks @emilymbender & @alex for doing #MAIHT3k.

Last week you helped me during a conversation with a colleague when talking about #AI to get him to see a new perspective on #AI using the points you've advocated and praticed in #MAIHT3k:

• Nobody should be subjected to synthetic text without their knowledge & consent.
• Synthetic text has no meaning, no intention
• LLMs are not people and shouldn't be treated as such (grammatically, in academic literature or elsewhere).
• LLMs are produced on the serial abuse of workers for labelling data and censoring outputs.
• LLMs are built on the serial abuse of authors and artists.
• LLMs have unknown, but enormous, environmental impacts.

We work at an educational insitute that'll organize two weeks on #AI aimed at teachers and high school students in a few months and I hope that some of the critical perspective you have presented in #MAIHT3k might influence it.

csdummi, 4 months ago

@liaizon @bookwyrm @hugh @tripofmice @mattlehrer

It's great to know the work is appreciated!

Thanks to @hugh and @tripofmice for getting this project merged.

The development of this feature was a team effort organized as a #GuildAlpha sprint and would not have been possible without the help of @slumberingcat, @tomat0, @RyunoKi and @smallcircles among others.

csdummi, 4 months ago

@hausaffe

And for #Fediverse related talks:

Who Killed The Internet? And a promising alternative for Public Communication and Social Media: the Fediverse!

https://events.ccc.de/congress/2023/hub/event/who-killed-the-internet-and-a-promising-alternative-for-public-communication-and-social-media-the-fediverse/

Though the title js a bit misleading: it mainly focuses on #EUVoice, #EUVideo - and the problems and concerns that governments have (and should have) with publishing and using social media (especially such that they don't own).

#37c3 #fahrplan