CoAStAL NLP

Research collective at University of Copenhagen interested in all things NLP and a ton of things not NLP.

[Laura Cabello @LauraCabelloPi]

Hey our work on ChatGPT was featured on @borsendk! 🗞️ "New study shows that the popular chatbot ChatGPT promotes American norms and values. Scientists warn against using it blindly." Here it's a nice shot 📸 of the printed version. Together w/ @daniel_hers @coastalcph

[CoAStAL NLP @coastalcph]

New 1y+6m COASTAL postdoc on trustworthy NLU around claims/facts. Reach out to [email protected] if interested. #nlproc #ai #LLMs

[CoAStAL NLP @coastalcph]

@yuxiangw_cs Strictly speaking, the claim is non-sensical, since you cannot sample at random from human-written text (because poorly defined and subject to drift).

[CoAStAL NLP @coastalcph]

@yoavartzi https://t.co/AnJSWJ4Bm6

[Miryam de Lhoneux/ @mdlhx.bsky.social @mdlhx]

I am hiring two PhD students in multilingual #nlproc based @CW_KULeuven! Info and application link: https://t.co/REa6hIMKEx Deadline 8/09. RTs and sharing on other social media appreciated! Note that I am on parental leave and therefore slow at answering emails/DMs etc.

[Andrew Ilyas @andrew_ilyas]

What makes ML models tick? How do we attribute model behavior to the training data, algorithm, architecture, or scale used in training?

Papers (or ideas) here? Submit to ATTRIB @ NeurIPS 2023 (https://t.co/BVVrBj5jjm)!

Deadline is September 23!

[Desmond Elliott @delliott]

I'm hiring a postdoc for 18 months to work with us on tokenization-free NLP in Copenhagen. Applications are due by August 31: https://t.co/Bbefd9jxij. Don't hesitate to reach out with informal inquiries.

[UKP Lab @UKPLab]

At 11:00 (17:00 CET) in the Frontenac Ballroom and Queen's Quay: »What does the Failure to Reason with ›Respectively‹ in Zero/Few-Shot Settings Tell Us about Language Models?« by @ruixiangcui, @Seolhwa_Lee_, @daniel_hers and Anders Søgaard.

https://t.co/143eVdjXwT #ACL2023NLP

[CoAStAL NLP @coastalcph]

@glupyan See also: https://t.co/VrdXzjMYtV

[Markus Giessing @GiessingMarkus]

I denne uges @weekendavisen skriver jeg om "tankelæsninger", hvilket jeg har forsøgt at blive klogere på med hjælp fra Anders Søgaard (KU) og @GretaTuckute (M.I.T.), der begge forsker i det kuriøse emne.

[Rita Ramos @RitaRamos1]

We're presenting SmallCap, a lightweight image captioning model with retrieval augmentation at @CVPR. If you're around, come chat with us! #CVPR2023 @YovaKem_v2

[Andreas Roepstorff @androrff]

Vil du hurtigt opdateres begavet på, hvad der sker i AI feltet, hvorfor det er vigtigt, og hvordan vi kan begynde at navigere i det? Så begynd her, @RebAdlerNissen, @suneman @M_B_Petersen og Anders Søgaard i skarp kronik @politiken.

[Carlsbergfondet @Carlsbergfondet]

Anders Søgaard: Strømforbrug er en enorm udfordring ift kunstig intelligens. Strømforbruget i et af de første træningsrul af chatGPT svarede til et års husholdningsstrømforbrug. Heldigvis bliver tech virk hhv bedre til at mindre strøm + investerer i vedvarende energi #fmdk #dkforsk

[Shriram Krishnamurthi (primary: Bluesky) @ShriramKMurthi]

Very sad to announce that Eugene Charniak, a statistical NLP pioneer ("Charniak parsing"), inaugural class ACL fellow, etc., passed away this morning.

Here's an article written at his retirement, looking back on his career.

https://t.co/ljUYoea90R

[Rebecca Adler-Nissen @RebAdlerNissen]

I vores 3. kronik i @politiken om kunstig intelligens ser vi på den globale magtkoncentration. Det er ikke kun et kapløb om penge og data, men om kontrol over vores tilgang til viden, som sprogmodeller og chatbots genererer om os https://t.co/6bOPyj6I1Y

[Rita Ramos @RitaRamos1]

Multilingual image captioning models are usually trained with large-scale machine translated data.

In our ACL 2023 Findings paper https://t.co/mxTAxBL3IH, we propose LMCap, a few-shot and retrieval augmented captioning model that works without any task-specific training.

[Terne Sasha Thorn Jakobsen @Ternethorn]

Are you interested in #NLProc explainability and fairness? Then you’re in luck - the preprint for our #ACL2023 paper is out! https://t.co/GNtbadstmd Work done with the lovely @LauraCabelloPi and Anders Søgaard. Summary 🧵

[Laura Cabello @LauraCabelloPi]

Our paper "Being Right for Whose Right Reasons?" is now available as a preprint on arXiv 🐣

[Kforum @Kforum]

Hvad gør mennesker til noget særligt, når kunstig intelligens kan det samme som os? 

Anders Søgaard, professor i datalogi og filosofi ved Københavns Universitet, mener, at kunstig intelligens vil forandre verden på måder, vi måske slet ikke forstår.

https://t.co/whMZTcGt0I

[Ilias Chalkidis @KiddoThe2B]

💡Our paper "Efficient Document Embeddings via Self-Contrastive Bregman Divergence Learning" with Daniel Saggau, and @MinaRezaei11 from @LMU_Muenchen has been accepted at #ACL2023NLP Findings. Preprint 📚: https://t.co/WH9F7Ey8vr 🧵👇 1/N