Conversation

Notices

1. Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 22:52:23 UTC Bernie

   I'm watching the livestream of Aaron Swartz Day 2022
   https://www.aaronswartzday.org/asd-2022-livestream/

   #aaronswartz #openaccess #hackathon #privacy

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:00:29 UTC Bernie

     in reply to

     I can't find the schedule for today's livestream...

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:12:46 UTC Bernie

     in reply to

     Cindy Cohn, director of the #EFF, is speaking now 👏

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:18:41 UTC Bernie

     in reply to

     The EFF stands out among other advocacy organizations because their communication is always highly factual and free from hyperbolic analogies and FUD.

     Just now, Cohn pointed out that Google does not, in fact, sell user data. Google *monetizes* user data. And she explained the difference.

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:23:04 UTC Bernie

     in reply to

     When criticizing Google (or any other business), it's important to be factually correct, or else they can simply shrug and say "no, we're not selling your data and never will".

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:34:53 UTC Bernie

     in reply to

     Not long ago, I defended Microsoft / GitHub because people where saying that Copilot violates opensource licenses or even "steals code".

     It's not a given that a model trained on opensource code will generate copyright infringing output. If it does, it should be easy to show a couple of examples as evidence. Right?

     #copyright #eff

   • Luis/Luigi (luigidev@mastodon.lol)'s status on Saturday, 12-Nov-2022 23:55:01 UTC Luis/Luigi

     in reply to

     @codewiz it's been shown to output verbatim pieces of code actually. Like actually copy-pasting whole functions

   • Bernie (codewiz@mstdn.io)'s status on Saturday, 12-Nov-2022 23:55:32 UTC Bernie

     in reply to

     • Luis/Luigi

     @LuigiDev Oh, it still does that?

     It used to in July 2021, but that was a long time ago:
     https://news.ycombinator.com/item?id=27710287

     If this is still the state of the art, then it's not only copyright infringing, it's also a disappointingly bad result for a well funded team of ML experts 😂

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 00:00:56 UTC Bernie

     in reply to

     • Luis/Luigi

     @LuigiDev Oh, it's still doing that today?

     It used to in July 2021, but that was a long time ago in this rapidly advancing field:
     https://news.ycombinator.com/item?id=27710287

     Not only this is copyright infringement plain outright, it would also be a disappointingly bad result for a well funded team of ML researches 😂

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 00:30:36 UTC Bernie

     in reply to

     • Luis/Luigi

     @LuigiDev Are they trying to sell a tool which is no better than a student copy-pasting the wrong answer from Stack Exchange and getting caught because they can't even rename the variables?

   • Luis/Luigi (luigidev@mastodon.lol)'s status on Sunday, 13-Nov-2022 00:30:37 UTC Luis/Luigi

     in reply to

     @codewiz I mean the Microsoft proposed solution is to compare the copilot output against GitHub and if there's an exact match then abort it. That doesn't give a lot of hope really.

     And what is worse, this "fix" is opt-in

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 00:49:59 UTC Bernie

     in reply to

     Now speaking: Mark Graham, director of the Wayback Machine 👏
     https://www.youtube.com/watch?v=BWfqV_adW54

     #InternetArchive #aaronswartzday

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 01:15:05 UTC Bernie

     in reply to

     Wikipedia and the Internet Archive are amazing organizations.

     For decades, they've been pursuing incredibly ambitious missions, entirely funded by donations and unimpaired by the greed of shareholders and politicians.

     Their operating budget is tiny; the services delivered to humanity are priceless.

     #Wikipedia #WikiMedia #InternetArchive #nonprofit #internet

   • Michael Downey :xr: (downey@floss.social)'s status on Sunday, 13-Nov-2022 01:44:03 UTC Michael Downey :xr:

     in reply to

     @codewiz #Wikipedia #WikiMedia

     https://www.dailydot.com/debug/wikipedia-endownemnt-fundraising/

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 01:44:41 UTC Bernie

     in reply to

     • Michael Downey :xr:

     @downey Exactly. I'm aware that Wikipedia has more funding that they require.

     Perhaps surprisingly, crowd-sourcing all the knowledge of humanity, in dozens of different languages and making it freely available worldwide...all for a tiny fraction of the resources allocated to relatively simpler platforms for microblogging (*) and instant messaging (**).

     (*) Twitter sold for $44B
     (**) WhatsApp sold for $19B in 2014

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 01:56:42 UTC Bernie

     in reply to

     • Michael Downey :xr:

     Before someone corrects me, I should point out that the valuation of Twitter and WhatsApp isn't strictly correlated to their operating costs or their complexity.

     Everyone here knows that a platform equivalent to Twitter could be redesigned by a small group of opensource contributors and operated by a handful of volunteer sysadmins 😁

     @downey

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 02:02:34 UTC Bernie

     in reply to

     • Michael Downey :xr:

     The market value of Twitter and WhatsApp comes from their established user bases, and from the expectation that those users will somehow remain loyal when aggressively monetized...

     Well, good luck! 🦄

     @downey

   • Bernie (codewiz@mstdn.io)'s status on Sunday, 13-Nov-2022 02:24:13 UTC Bernie

     in reply to

     • Michael Downey :xr:

     @downey Federating Wikipedia on multiple domains scattered around the world wouldn't really make encyclopedic knowledge more available or more useful.

     You can already link across wikis using standards like InterWiki, but ultimately plain web urls work just fine.

     Perhaps more resilient to government interference? But the entire corpus of Wikipedia is officially available for download, and mirrors are allowed.

   • Michael Downey :xr: (downey@floss.social)'s status on Sunday, 13-Nov-2022 02:24:14 UTC Michael Downey :xr:

     in reply to

     @codewiz

     As a collection of human knowledge #Wikipedia should be liberated using Ward Cunningham's federated wiki concept. Of course then the #WMF would be unnecessary and would lose that sweet sweet cash so not ever gonna be something they promote.

     #nonprofits

   • Youth (micheldebree@mastodon.social)'s status on Wednesday, 16-Nov-2022 09:03:26 UTC Youth

     in reply to

     @codewiz I read that they have millions and millions of cash, and only need a fraction of that to operate. https://unherd.com/thepost/the-next-time-wikipedia-asks-for-a-donation-ignore-it/ I don't know what to believe now.

   • Bernie (codewiz@mstdn.io)'s status on Wednesday, 16-Nov-2022 09:03:55 UTC Bernie

     in reply to

     • Youth

     @micheldebree Millions of dollars *is* a tiny budget for one of the top-7 websites. The others have revenues on the scale of billions.

     https://en.wikipedia.org/wiki/List_of_most_visited_websites