The below is an off-site archive of all tweets posted by @texttheater ever

Recent tweets

spinfocl Ich möchte die wenigsten der veröffentlichten Texte und Tweets lesen, aber hey, das hier ist kein Ponyhof. twitter.com/antonrainer/st…

via Twitter for iPad (retweeted on 11:38 PM, Dec 2nd, 2021 via Twitter Web App)

Unvollständige Liste deutschsprachiger Filme, in denen die Dialogzeile „Sag mal, spinnst du?“ vorkommt.

Toni Erdmann
Fucking Berlin
Mein Bruder heißt Robert und ist ein Idiot („Sag mal, spinnst… mff“)
The Trouble with Being Born („Sag amoi, spinnst?“)

TBC

via Twitter Web App

zeitweise So jetzt Hot Take:
Jede gute Spiel-, Buch-, Filmkritik braucht auch einen ausführlichen Spoilerteil.

via Twitter for iPhone (retweeted on 10:05 PM, Dec 2nd, 2021 via Twitter for Android)

hllizi (see below the above the above for the above)

via Twitter for Android (retweeted on 8:59 PM, Dec 2nd, 2021 via Twitter for Android)

GabrielBerlin BREAKING: German chancellor Merkel to have Farbfilm vergessen at her very own zap strike

via Twitter for Mac (retweeted on 8:57 PM, Dec 2nd, 2021 via Twitter for Android)

frank_opitz Wir schicken die Bundeswehr nächstes Jahr zum ESC.

via Twitter for iPhone (retweeted on 8:56 PM, Dec 2nd, 2021 via Twitter for Android)

beispielwiese Selbstbild als Bruttoregistertonne.

via Twitter for Android (retweeted on 8:56 PM, Dec 2nd, 2021 via Twitter for Android)

formschub Ich kann übrigens gleichzeitig für Meinungsfreiheit sein und in meiner Fantasie trotzdem den geheimen Wunsch hegen, einigen medial überpräsenten Pandemietreibern die Fresse zuzutackern,

via Twitter for iPhone (retweeted on 8:54 PM, Dec 2nd, 2021 via Twitter for Android)

MLKessel Not only is today's date a palindrome, it's also an ambigram. You can read it left to right, right to left, and also upside down pic.twitter.com/qX43dIUHoI

via Twitter Web App (retweeted on 8:44 PM, Dec 2nd, 2021 via Twitter for Android)

hllizi zweimal kurz kanzler

via Twitter for Android (retweeted on 8:42 PM, Dec 2nd, 2021 via Twitter for Android)

Laserhedvig Glottolog editors Harald Hammarström & @xrotwang have written a paper explaining how glottocodes work and their history. essential information to anyone doing cross-linguistic work!
semantic-web-journal.net/content/glotto…

via Twitter Web App (retweeted on 8:14 PM, Dec 2nd, 2021 via Twitter for Android)

NYTScience This simulation provides an unprecedented glimpse at how the coronavirus survives in the open air as it spreads to a new host nyti.ms/3DaQrVx

via SocialFlow (retweeted on 7:47 PM, Dec 2nd, 2021 via Twitter for Android)

NewsfromScience Since South African scientists announced that they had identified an unsettling new variant of , many are wondering where and how the evolved—and what lessons its emergence holds for avoiding future dangerous variants. fcld.ly/g13qzuo pic.twitter.com/vvFGbrzXTe

via Facelift-Cloud (retweeted on 7:37 PM, Dec 2nd, 2021 via Twitter for Android)

washingtonpost Google fired star AI researcher Timnit Gebru one year ago. Now she’s launching her own institute. wapo.st/31nhHD0

via SocialFlow (retweeted on 7:31 PM, Dec 2nd, 2021 via Twitter for Android)

Wort des Tages: Gehörfindungsstrategie

via Twitter for Android

Eine nicht ganz so fatale, aber verwandte Design-Unart: die Bedienknöpfe von Küchenwaagen in die Wägefläche zu integrieren.

via Twitter for Android in reply to texttheater

“Oddly specific remark, must have been a reviewer question.”

via Twitter for Android

@matthiaskrgr1 Es kommt vor, dass es funktioniert, während letzteres nicht funktioniert. Andersrum ist aber häufiger.

via Twitter Web App in reply to matthiaskrgr1

sleepinyourhat Not really. It's important that we build a clear consensus on what kind of progress we are and aren't making in NLP. That means avoiding hype, but it also means acknowledging the progress that we do make.

via Twitter Web App (retweeted on 3:47 PM, Dec 2nd, 2021 via Twitter Web App)

sleepinyourhat I even put together a silly inspired rule of thumb: pic.twitter.com/UBeEWTt1mH

via Twitter Web App (retweeted on 3:46 PM, Dec 2nd, 2021 via Twitter Web App)

sleepinyourhat The argument in this paper is simple: The stakes are getting higher for what we do. We should be careful about the claims we make. Putting a sloppy or unjustified claim in a paper is a problem whether that claim is about what we can do or what we can't.

via Twitter Web App (retweeted on 3:46 PM, Dec 2nd, 2021 via Twitter Web App)

sleepinyourhat That kind of claim masks the huge difference between the small trained-from-scratch LSTMs of 2017 and newer models like GPT-3 or FLAN that barely use labeled training data at all. Newer systems are far less brittle, and they're not brittle in the same ways.

via Twitter Web App (retweeted on 3:46 PM, Dec 2nd, 2021 via Twitter Web App)

sleepinyourhat I see a lot of authors in NLP lately who seem to be going far out of their way to write that current systems don't really work well.

via Twitter Web App (retweeted on 3:45 PM, Dec 2nd, 2021 via Twitter Web App)