Cognition, language, literature, occasional possums. Elements of Surprise: Our Mental Limits and the Satisfactions of Plot. https://xitter.cf/t.co/7odHdeNphA
she/her/hers
Joined April 2017
- Tweets1.5K
- Following251
- Followers397
- Likes13.1K
Pinned Tweet
Want to read Elements of Surprise, but worried that it will blindside you with a spoiler for the book you were about to read or film you were about to see? I've got you covered. veratobin.org/publications/e…
Vera Tobin retweeted
THIS IS UNFATHOMABLY SICK. Wait til Chao Tian comes in, it goes so hard. New favorite genre: Appalachinese!!! 🪕🇨🇳
Vera Tobin retweeted
My lovely Cognitive Science department is hiring for a Visiting Assistant Professor position, starting January 2024 and ending December 2024. 2/2 teaching load (our undergrads are WONDERFUL), renewable for a second year. apply.interfolio.com/131626
My lovely Cognitive Science department is hiring for a Visiting Assistant Professor position, starting January 2024 and ending December 2024. 2/2 teaching load (our undergrads are WONDERFUL), renewable for a second year. apply.interfolio.com/131626
someone heard "little kitty, big city" but got confused @LittleKittyGame
ALT my big kitty relaxing in the little town square of a Lego city
This makes me so sad. Also, do follow Alariko for his gorgeous, charming art.
Amazing thread on Reddit about the AI model that used my art without my permission to train their model and even named it after me. Apparently I am the bad person for wanting to get it taken down, and wanting to “brigade” (??) it.
reddit.com/r/StableDiffusion…
don't need to see Oppenheimer, just need the water and the well
ALT three black and white screen caps from Twin Peaks: The Return part 8: the Woodsman, reflected in a rearview mirror; the Trinity nuclear test; a creature crawling into the mouth of a young woman asleep in her flowered nightgown
Vera Tobin retweeted
Ever wondered how nondeterministic GPT-4 is even with greedy decoding (T=0)? I built a website that asks GPT-4 to draw a unicorn every hour and tracks if the results stay consistent over time (spoiler alert: they don't! 🦄).
Explore the findings:
openaiwatch.com
Frankly delighted when @AlisonGopnik told everyone at this #acl2023nlp plenary talk to go read @henryfarrell
Vera Tobin retweeted
Please do some crosswords for science! iliad.stanford.edu/experimen…
We're studying how people solve xwords, and we are looking for solvers of all experience levels: from novices to ACPTers. High-quality puzzles generously provided by the wonderful boswords.org/ team.
Vera Tobin retweeted
We disentangle such overfitting from LMs' transferrable reasoning skills by introducing "counterfactual" variants of familiar tasks. Some of these, e.g. the drawing & music tasks, are not standard LM evaluations, but recent work has qualitatively observed these abilities in LMs.
An enjoyable talk about a new approach to some of the same kinds of analysis we often do with Word2Vec and the like, with a focus on cases of semantic change (something of special interest to me right now). Interested to see where things go with this work!
Exciting collaboration with @northernkender, Iris Luden (MSc Logic student in Amsterdam🚨), and @raquel_dmg
Looking forward to discussing the paper both before and during #ACL2023NLP :)
Someone asks a good question about how this introduces "subjectivity" (or fuzziness, variability, black box factor), detracting from the attractive "objectivity" of word embeddings. Would love a tool that exposes underlying clusters together with such qualitative representations
Oops, PS: these tweets about arxiv.org/abs/2305.11993 should have included #ACL2023NLP
Intriguing prospects! Looking forward to reading these.
Are you a big fan of structure?
Have you ever wanted to apply the latest and greatest large language model out-of-the-box to parsing?
Are you a secret connoisseur of linear-time dynamic programs?
If you answered yes, our outstanding #ACL2023NLP paper may be just right for you!
Very enjoyable and crunchy dive into the gaps, imbalances, and other issues in comparing human and machine performance on various tasks, with a very engaging presentation too, at #ACL2023NLP
📢 NEW #ACL2023 PAPER
"What's the meaning of superhuman performance in today's NLU?" 📊
w/ @JohanBos @ThierryDeclerck @HajicJan @daniel_hers @ehovy @alkoller @SimonKrek @StevenSchockae2 @RicoSennrich @KatiaShutova @RNavigli @Babelscape #NLProc
📑 arxiv.org/abs/2305.08414
Vera Tobin retweeted
Comparing, even equating, LLMs' language understanding with the understanding of a language-impaired person is not something I would like to see more of.
Geoffrey Hinton arguing that similarities between human and LLM performance should convince us that LLMs understand things; as a linguist, I think the better way to update our priors is to recognize that a lot of our own language use is in fact spicy autocomplete #ACL2023NLP
🗓️ Wednesday 09:00-10:30
📍Frontenac Ballroom and Queen’s Quay
#ACL2023NLP paper led by @AdityaYadavall2 and @TheAlekhya Modeling second language acquisition teaches us about negative transfer & child-directed speech. Also: a new multi-lingual CDS corpus! arxiv.org/abs/2305.19589