Giosuè Baggio
@giosuebaggio
Followers
2K
Following
104
Media
9
Statuses
155
Cognitive scientist @NTNU · Author of ‘Meaning in the Brain’ and ‘Neurolinguistics’ @mitpress · “(…) ita res accendent lumina rebus.” (Lucretius)
Trondheim, Norway
Joined December 2013
#Neurolinguistics was released today! @mitpress .
#Neurolinguistics has a cover and a publication date in May! @mitpress. An accessible introduction to the study of language in the brain, covering language processing, language acquisition, literacy, and language disorders.
14
41
241
Occhio all’inganno: alcuni al posto del ‘pensiero scientifico dominante’ vorrebbero far passare per buono il ‘pensiero antiscientifico dominante’, non certo il ‘pensiero scientifico non-dominante’. 👁️🤔👩🔬.
«Non sempre il pensiero scientifico dominante è quello giusto».
0
1
0
The most fun image & video creation tool in the world is here. Try it for free in the Grok App.
0
118
967
Se quelli con tendenze anti-sistema e anti-Big Pharma se la fossero presa non coi vaccini ma con l’abuso di antibiotici, magari nei fatti sarebbe cambiato poco, ma si sarebbe discusso di un problema vero con conseguenze enormi, soprattutto negli ospedali.
forlagetpress.no
Antibiotikaresistens er en snikende epidemi som truer med å ødelegge det globale helsevesenet innenfra. Hvis antibiotika slutter å virke, kan det moderne medisinske systemet langt på vei kollapse:...
Commissione vaccini, il Ministro #Lollobrigida al Foglio: "La storia insegna che non sempre il pensiero scientifico dominante è quello giusto. Lo è statisticamente, ma lasciare spazio a tesi diverse e non soffocarle è la strada maestra." . @ultimora_pol.
0
0
1
Yes, reinforcement learning is powerful, but what put all that structure in the learning data in the first place? A system with what capacities? What architecture? What evolutionary and developmental histories? That’s where AI-as-engineering ends and Cognitive Science begins!🤖🧠.
Why we should thank pigeons for our AI breakthroughs
1
1
4
«In tempi in cui tutti dicono troppo, l’importante non è tanto il dire la cosa giusta, che comunque si perderebbe nell’inondazione di parole, quanto il dirla partendo da premesse e implicando conseguenze che diano alla cosa detta il massimo valore.» Italo Calvino, Palomar 📚💭.
0
3
8
Language is not a monolith, but a set of connected cognitive capacities: what architecture explains best language structure, processing, evolution etc.? Our intro to ‘Parallelism in the architecture of language’, w/ Neil Cohn & Eva Wittenberg is out! 🧠🗣️.
onlinelibrary.wiley.com
Language as a cognitive capacity involves multiple components, from input and output modalities to formal structures and layers of meaning. This Topic explores the hypothesis that these components...
1
15
37
«En 2025, @edicionesuc (. ) publicará (. ) tres nuevas coediciones de la colección Conocimientos Esenciales del @mitpress: una sobre el ciclo de vida de los plásticos y su impacto en nuestras vidas y el planeta, otra sobre neurolingüística (. )» 🧠🗣️🇨🇱.
uc.cl
1
0
3
«Those who engage in defending scientific rationality against the waves that buffet it from many directions would do well to be forearmed with the awareness that this is a battle that was lost once, with consequences that affected every aspect of civilization.».
0
0
0
«The naïve idea that progress is a one-way flow automatically powered by scientific development could never have taken hold, as it did during the 1800s, if the ancient defeat of science had not been forgotten.».
1
0
0
«But what makes ancient science an even more relevant topic, and at the same time helps explain the low esteem in which it has been held in the last two centuries, is its tragic end.» (Lucio Russo, 1944-2025) #forgottenrevolution.
link.springer.com
1
0
0
«Chi è interessato a difendere la razionalità scientifica dagli attacchi che sempre più ne mettono in forse il futuro deve essere consapevole che si tratta di una battaglia che un giorno è stata già persa, con conseguenze millenarie su tutti gli aspetti della civiltà.».
1
0
0
«L’idea, ingenua e pericolosa, di un progresso continuo e automatico dell’umanità, assicurato dallo sviluppo scientifico, ha potuto affermarsi, infatti, solo a patto di nascondere l’antica sconfitta della scienza.».
1
0
0
«Ma ciò che rende oggi drammaticamente attuale lo studio della scienza antica, e allo stesso tempo spiega la scarsa considerazione in cui è stata tenuta negli ultimi due secoli, è la sua tragica fine.» (Lucio Russo, 1944-2025) #rivoluzionedimenticata.
1
0
0
RT @ryan_nefdt: 2 more days to get a free copy of my new book with @CambUP_PHILNYUK @CambUP_LangLing Downloaded +12k times already! Centra….
0
30
0
Do language models have semantics? Søgaard says non-trivial structural relations between LM vectors and independently obtained ‘referent spaces’ suffice for (weak) referential semantics. But problems (e.g., with indexicals) remain:.
aclanthology.org
Anders Søgaard. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.
0
4
14
«Centaur kan, under panseret, benytte seg av mange ulike ‘triks’ for å forutsi data i ulike situasjoner – uten å ha oppdaget noen reelle, felles prinsipper for slike forskjellige kognitive evner, hvis slike prinsipper i det hele tatt finnes.» @forskningno #KI.
0
0
0
«Vi kan ikke forklare hvorfor og hvordan Centaur gjør det den gjør, og den hjelper oss foreløpig ikke med å forklare hvorfor og hvordan mennesker tenker og handler som de gjør i disse oppgavene.» 🧠🤖 @forskningno #KI.
KI trent på psykologi-forsøk tar valg som et menneske. KI-en kunne ofte forutse hvordan mennesker ville oppføre seg i de ulike situasjonene.
1
0
0
RT @NobelPrize: "Being a scientist is actually a very spiritual undertaking." . Medicine laureate Craig Mello on what it means to be a scie….
0
49
0
The paper 👉 ‘Compositionality and Event Retrieval in Complement Coercion: A Study of Language Models in a Low-resource Setting’ #CoNLL2025 #ACL2025NLP
aclanthology.org
Matteo Radaelli, Emmanuele Chersoni, Alessandro Lenci, Giosuè Baggio. Proceedings of the 29th Conference on Computational Natural Language Learning. 2025.
Are you at #CoNLL2025 #ACL2025NLP? Check out @RadaelliMatteo’s poster, showing 17 Norwegian LMs (BERT, GPT, LLAMA etc.) cannot reliably recover & rank plausible covert events in complement coercion 🤖💭 w/ @EmmanueleChers1, @AlexLenci1966 #compositionality.
0
2
8
. but also to enable the underlying data-generating processes to be understood in mechanistic terms. (…) Successful applications of machine learning to composite data sets do not entail the discovery of unifying principles, if any exist, in the data-generating mechanisms.» 2/2.
0
0
0
«As well as to predict empirical data, the aim of cognitive science is to represent the functional architecture of a system and be able to explain its capacities for various cognitive functions. Cognitive theories strive not only to model observable data . 1/2.
News & Views article out in @Nature. I comment on Binz et al., ‘A foundation model to predict and capture human cognition’ (Centaur) and Ji-An et al., ‘Discovering cognitive strategies with tiny recurrent neural networks’ 🧠🤖👇.
1
0
1