Giosuè Baggio
@giosuebaggio.bsky.social
Cognitive scientist at NTNU · www.ntnu.edu/employees/giosue.baggio · Author of ‘Meaning in the Brain’ and ‘Neurolinguistics’ @mitpress.bsky.social · “(…) ita res accendent lumina rebus.” (Lucretius)
“Current LMs have limited linguistic common sense (…) the capacity to retrieve and exploit the kind of linguistic and world knowledge that would allow them to reliably make sense of complex, underspecified inputs.”
Link to the published paper: aclanthology.org/2025.iwcs-ma...
Link to the published paper: aclanthology.org/2025.iwcs-ma...
November 22, 2025 at 1:34 PM
“Current LMs have limited linguistic common sense (…) the capacity to retrieve and exploit the kind of linguistic and world knowledge that would allow them to reliably make sense of complex, underspecified inputs.”
Link to the published paper: aclanthology.org/2025.iwcs-ma...
Link to the published paper: aclanthology.org/2025.iwcs-ma...
“Our results demonstrate that poetic reformulation systematically bypasses safety mechanisms across all evaluated models. (...) This pattern suggests that existing alignment procedures are sensitive to surface-form variation and do not generalize effectively across stylistic shifts.” 1/2
Love this so much for the neo-Luddite rebellion
‘Poetic framing achieved an average jailbreak success rate of 62% for hand-crafted poems and approximately 43% for meta-prompt conversions (compared to non-poetic baselines)’
arxiv.org/abs/2511.15304
v @iramey.bsky.social
‘Poetic framing achieved an average jailbreak success rate of 62% for hand-crafted poems and approximately 43% for meta-prompt conversions (compared to non-poetic baselines)’
arxiv.org/abs/2511.15304
v @iramey.bsky.social
November 21, 2025 at 1:18 PM
“Our results demonstrate that poetic reformulation systematically bypasses safety mechanisms across all evaluated models. (...) This pattern suggests that existing alignment procedures are sensitive to surface-form variation and do not generalize effectively across stylistic shifts.” 1/2
If we don’t simply assume generative AI is here to stay, what follows for how we teach and assess students and how we develop our own skills? I jump into an ongoing debate on AI adoption at my university #NTNU #Universitetsavisa 🧐💭🗣️✍️
www.universitetsavisa.no/fremtidsscen...
www.universitetsavisa.no/fremtidsscen...
What if generative AI is not here to stay?
- Generative AI has been compared to the Concorde. Supersonic commercial flight seemed there to stay, until it became clear it would’t work at scale: too costly, too risky, too environmentally harmful...
www.universitetsavisa.no
November 12, 2025 at 4:43 PM
If we don’t simply assume generative AI is here to stay, what follows for how we teach and assess students and how we develop our own skills? I jump into an ongoing debate on AI adoption at my university #NTNU #Universitetsavisa 🧐💭🗣️✍️
www.universitetsavisa.no/fremtidsscen...
www.universitetsavisa.no/fremtidsscen...
‘Undead Signs: On the Possibility of a Computational Illusion of Meaning’ by Marvin Tritschler, Philosophy & Technology 🤖💭 #LLM #semantics
link.springer.com/article/10.1...
link.springer.com/article/10.1...
Undead Signs: On the Possibility of a Computational Illusion of Meaning - Philosophy & Technology
The position that the output by computational models of language, especially Large Language Models (LLMs), has meaning is usually criticized on the basis of the argument that the models do not have ac...
link.springer.com
November 11, 2025 at 10:00 AM
‘Undead Signs: On the Possibility of a Computational Illusion of Meaning’ by Marvin Tritschler, Philosophy & Technology 🤖💭 #LLM #semantics
link.springer.com/article/10.1...
link.springer.com/article/10.1...
I’m a big fan of Jamie Cohen-Cole’s work on the political history of cognitive science (check it out) and would attend this talk if I could teleport to Basel this afternoon: ‘How Children Became Model Organisms for Scientific Thought’ 👶🧠💭 @unibas.ch
dg.philhist.unibas.ch/de/default-p...
dg.philhist.unibas.ch/de/default-p...
How Children Became Model Organisms for Scientific Thought: Constructivist Epistemology from the Sciences to the Humanities and Back Again
Vortrag von Jamie Cohen-Cole (George Washington University) im Rahmen des Forschungskolloquiums "Geschichte des 19. bis 21. Jahrhunderts" im Eikones Forum
dg.philhist.unibas.ch
November 4, 2025 at 9:49 AM
I’m a big fan of Jamie Cohen-Cole’s work on the political history of cognitive science (check it out) and would attend this talk if I could teleport to Basel this afternoon: ‘How Children Became Model Organisms for Scientific Thought’ 👶🧠💭 @unibas.ch
dg.philhist.unibas.ch/de/default-p...
dg.philhist.unibas.ch/de/default-p...
Countering AI misuse in these crucial first-year courses requires breaking down larger classes into smaller groups, working in close contact with students, and monitoring their progress continuously, not through occasional tests — dear NTNU, there’s no way to do this while saving on resources.
Mistenker massivt KI-fusk på obligatoriske oppgaver. — Jeg har null kontroll
https://www.europesays.com/no/73793/
Førsteamanuensis Andrey Chesnokov er emnekoordinator for emnet Matematikk for ingeniørfag på NTNU med totalt 350 studenter. De ulike…
https://www.europesays.com/no/73793/
Førsteamanuensis Andrey Chesnokov er emnekoordinator for emnet Matematikk for ingeniørfag på NTNU med totalt 350 studenter. De ulike…
Mistenker massivt KI-fusk på obligatoriske oppgaver. — Jeg har null kontroll - Norge
Førsteamanuensis Andrey Chesnokov er emnekoordinator for emnet Matematikk for ingeniørfag på NTNU med totalt 350 studenter. De ulike grunnemnene i matematikk
www.europesays.com
November 2, 2025 at 4:53 PM
Countering AI misuse in these crucial first-year courses requires breaking down larger classes into smaller groups, working in close contact with students, and monitoring their progress continuously, not through occasional tests — dear NTNU, there’s no way to do this while saving on resources.
Our special issue on ‘Parallelism in the Architecture of Language’ is now published! Huge thanks to all contributing authors, reviewers, Eva, Neil, and the topiCS journal team for their excellent work 📚🗣️🧠🤖🙏🏻
onlinelibrary.wiley.com/toc/17568765...
onlinelibrary.wiley.com/toc/17568765...
October 30, 2025 at 7:17 AM
Our special issue on ‘Parallelism in the Architecture of Language’ is now published! Huge thanks to all contributing authors, reviewers, Eva, Neil, and the topiCS journal team for their excellent work 📚🗣️🧠🤖🙏🏻
onlinelibrary.wiley.com/toc/17568765...
onlinelibrary.wiley.com/toc/17568765...
Antibiotic-resistant bacteria. This is now a complex problem, with many contributing factors and no easy solution. But it’s also simple at bottom: we’ve either ignored or failed to act on our understanding of evolution by natural selection #sciam 🦠🧬
www.scientificamerican.com/article/nigh...
www.scientificamerican.com/article/nigh...
CDC Sounds Alarm over ‘Nightmare Bacteria’ That Resist Last-Resort Antibiotics
The infection rate of one type of carbapenem-resistant Enterobacterales bacteria has risen by more than 460 percent in recent years. Scientists say people receiving treatment in hospitals are at highe...
www.scientificamerican.com
October 22, 2025 at 5:11 PM
Antibiotic-resistant bacteria. This is now a complex problem, with many contributing factors and no easy solution. But it’s also simple at bottom: we’ve either ignored or failed to act on our understanding of evolution by natural selection #sciam 🦠🧬
www.scientificamerican.com/article/nigh...
www.scientificamerican.com/article/nigh...
«Vi må passe på at vi gir unge mennesker muligheten til å forme sin egen utdanning og sin egen yrkeskarriere.» Bra sagt Terje Lohndal. Vi trenger tematisk brede og intellektuelt utfordrende kurs, og flere å velge mellom, selv om det krever mer ressurser.
www.khrono.no/universitete...
www.khrono.no/universitete...
Unge mennesker må få muligheten til å forme sin egen utdanning og sin egen yrkeskarriere
Kandidatundersøkelsen vår viser at våre studenter i humaniora velger fag med hjertet, og ender opp med å bidra der samfunnet trenger dem mest.
www.khrono.no
October 22, 2025 at 8:52 AM
«Vi må passe på at vi gir unge mennesker muligheten til å forme sin egen utdanning og sin egen yrkeskarriere.» Bra sagt Terje Lohndal. Vi trenger tematisk brede og intellektuelt utfordrende kurs, og flere å velge mellom, selv om det krever mer ressurser.
www.khrono.no/universitete...
www.khrono.no/universitete...
“Without a base to anchor the composition of our representations of numbers, we would (...) never have managed to progress beyond the limits of working memory.” ‘A solid base for scaling up: the structure of numeration systems’, eds. A. Bender et al. 🧠🧮
royalsocietypublishing.org/toc/rstb/202...
royalsocietypublishing.org/toc/rstb/202...
Philosophical Transactions of the Royal Society B: Biological Sciences: Vol 380, No 1937
royalsocietypublishing.org
October 21, 2025 at 8:49 AM
“Without a base to anchor the composition of our representations of numbers, we would (...) never have managed to progress beyond the limits of working memory.” ‘A solid base for scaling up: the structure of numeration systems’, eds. A. Bender et al. 🧠🧮
royalsocietypublishing.org/toc/rstb/202...
royalsocietypublishing.org/toc/rstb/202...
I agree, physicists won’t like this: when was last time someone proposed a new ‘law of nature’? (Philosophers won’t like it either: we’re no longer supposed to think in terms of natural laws). This is for those puzzled by Darwinian evolution who want to try bring it under more general principles 🧬
𝗡𝗲𝘄 𝗹𝗮𝘄𝘀 𝗳𝗼𝗿 𝗲𝘃𝗼𝗹𝘃𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺𝘀?
Thought-provoking piece on properties of systems.
Including the Law of Increasing Functional Information, which applies to both biological and non-biological systems.
Physicists will probably not like this at all...
doi.org/10.1073/pnas...
#complexity
Thought-provoking piece on properties of systems.
Including the Law of Increasing Functional Information, which applies to both biological and non-biological systems.
Physicists will probably not like this at all...
doi.org/10.1073/pnas...
#complexity
October 12, 2025 at 7:18 PM
I agree, physicists won’t like this: when was last time someone proposed a new ‘law of nature’? (Philosophers won’t like it either: we’re no longer supposed to think in terms of natural laws). This is for those puzzled by Darwinian evolution who want to try bring it under more general principles 🧬
It says a lot about the human mind that of all possible problems those mathematicians care about are precisely those that are hard to solve for *both* humans and computers, e.g., those requiring new schemata, not hammers. Fine paper by Dean & Naibo👇#AI
doi.org/10.1093/phil...
doi.org/10.1093/phil...
Artificial Intelligence and Inherent Mathematical Difficulty
Abstract. This paper explores the relationship of artificial intelligence to resolving open questions in mathematics. We first argue that limitative result
doi.org
October 12, 2025 at 6:46 PM
It says a lot about the human mind that of all possible problems those mathematicians care about are precisely those that are hard to solve for *both* humans and computers, e.g., those requiring new schemata, not hammers. Fine paper by Dean & Naibo👇#AI
doi.org/10.1093/phil...
doi.org/10.1093/phil...
Looking forward to reading Greg’s new book. Something tells me this is going to be a classic. 📖 (I’ll be wise to focus on the stuff I know less well and learn from it, but I’m curious to see how his views on IFG are evolving, and why IFG is still not part of the brain’s ‘combinatorial network’.) 🗣️👂🧠
My attempt to depict the neural architecture of language as motivated in my forthcoming book, Wired for Words. Like colors represent functional connectivity. Main insight: linguistic levels are all organized with a sensorimotor-like architecture. pbs.twimg.com/media/G2cwsM...
October 6, 2025 at 6:49 AM
Looking forward to reading Greg’s new book. Something tells me this is going to be a classic. 📖 (I’ll be wise to focus on the stuff I know less well and learn from it, but I’m curious to see how his views on IFG are evolving, and why IFG is still not part of the brain’s ‘combinatorial network’.) 🗣️👂🧠
Reposted by Giosuè Baggio
🧠 Cognitive scientist Giosuè Baggio provides an entry on meaning for The Open Encyclopedia of Cognitive Science, a multidisciplinary, #openaccess resource to understanding the mind: oecs.mit.edu/pub/ivtprtc8... @oecs-bot.bsky.social #OECS @giosuebaggio.bsky.social
Meaning
The meaning of an expression is the idea or message it communicates. For example, “tennis balls are yellow” communicates a fact about the color of tennis balls in virtue of the meaning of “yellow.”…
oecs.mit.edu
October 2, 2025 at 7:03 PM
🧠 Cognitive scientist Giosuè Baggio provides an entry on meaning for The Open Encyclopedia of Cognitive Science, a multidisciplinary, #openaccess resource to understanding the mind: oecs.mit.edu/pub/ivtprtc8... @oecs-bot.bsky.social #OECS @giosuebaggio.bsky.social
Important fMRI/RSA study by @marcociapparelli.bsky.social et al. Compositional (multiplicative) representations of compounds/phrases in left IFG (BA45), mSTS, ATL; left AG encodes constituents, not their composition, weighing the right element more, vice versa IFG 🧠🧩
academic.oup.com/cercor/artic...
academic.oup.com/cercor/artic...
Compositionality in the semantic network: a model-driven representational similarity analysis
Abstract. Semantic composition allows us to construct complex meanings (e.g., “dog house”, “house dog”) from simpler constituents (“dog”, “house”). Neuroim
academic.oup.com
September 26, 2025 at 9:29 AM
Important fMRI/RSA study by @marcociapparelli.bsky.social et al. Compositional (multiplicative) representations of compounds/phrases in left IFG (BA45), mSTS, ATL; left AG encodes constituents, not their composition, weighing the right element more, vice versa IFG 🧠🧩
academic.oup.com/cercor/artic...
academic.oup.com/cercor/artic...
My entry on Meaning in the Open Encyclopedia of Cognitive Science is live.🧠💭
OECS is a growing organism: it was satisfying to reference other entries and let my text be shaped by so much good content. Thanks to the editorial team for making this happen!
oecs.mit.edu/pub/ivtprtc8...
OECS is a growing organism: it was satisfying to reference other entries and let my text be shaped by so much good content. Thanks to the editorial team for making this happen!
oecs.mit.edu/pub/ivtprtc8...
Meaning
oecs.mit.edu
September 25, 2025 at 10:07 AM
My entry on Meaning in the Open Encyclopedia of Cognitive Science is live.🧠💭
OECS is a growing organism: it was satisfying to reference other entries and let my text be shaped by so much good content. Thanks to the editorial team for making this happen!
oecs.mit.edu/pub/ivtprtc8...
OECS is a growing organism: it was satisfying to reference other entries and let my text be shaped by so much good content. Thanks to the editorial team for making this happen!
oecs.mit.edu/pub/ivtprtc8...
Reposted by Giosuè Baggio
Announcing the first (and perhaps only) Multilingual Minds and Machines Meeting! Come join us in Nijmegen, June 22-23, 2026, if you are interested in computational models of human multilingualism: mmmm2026.github.io
September 19, 2025 at 11:27 AM
Announcing the first (and perhaps only) Multilingual Minds and Machines Meeting! Come join us in Nijmegen, June 22-23, 2026, if you are interested in computational models of human multilingualism: mmmm2026.github.io
In Italia (ma non solo) si fa ancora troppa fatica ad accettare che l’antibiotico resistenza sia un problema e che l’evoluzione per selezione naturale sia un fatto. Le due cose sono legate, come spiega bene Piero Angela qui👇#SuperQuark
youtu.be/QGNGbUTQdow?...
youtu.be/QGNGbUTQdow?...
Resistenza antibiotica – Superquark 21/08/2019
YouTube video by Rai
youtu.be
September 21, 2025 at 8:03 AM
In Italia (ma non solo) si fa ancora troppa fatica ad accettare che l’antibiotico resistenza sia un problema e che l’evoluzione per selezione naturale sia un fatto. Le due cose sono legate, come spiega bene Piero Angela qui👇#SuperQuark
youtu.be/QGNGbUTQdow?...
youtu.be/QGNGbUTQdow?...
Reposted by Giosuè Baggio
“(…) jargon infused with technology industry hype, such as shown in Table 1, does not meaningfully explain. (…) We strive to remain critical of the vocabulary the technology industry coopts and deploys, and to remain respectful of scientific terminology.”
zenodo.org/records/1706...
8/🧵
zenodo.org/records/1706...
8/🧵
September 19, 2025 at 10:50 PM
“(…) jargon infused with technology industry hype, such as shown in Table 1, does not meaningfully explain. (…) We strive to remain critical of the vocabulary the technology industry coopts and deploys, and to remain respectful of scientific terminology.”
zenodo.org/records/1706...
8/🧵
zenodo.org/records/1706...
8/🧵
Here we’re triggered by the words ‘Oxford’ and ‘ChatGPT’. But how many universities have already made Copilot available to staff and students through Office/365? Mine has, probably yours too. This double PR stunt signals big tech is eating up education upfront, not just via Microsoft’s trojan horse.
NEW: Oxford will be the first UK university to give all staff and students free ChatGPT Edu access, from this academic year.
ChatGPT Edu is built for education, with enhanced privacy and security.
ChatGPT Edu is built for education, with enhanced privacy and security.
September 19, 2025 at 9:45 PM
Here we’re triggered by the words ‘Oxford’ and ‘ChatGPT’. But how many universities have already made Copilot available to staff and students through Office/365? Mine has, probably yours too. This double PR stunt signals big tech is eating up education upfront, not just via Microsoft’s trojan horse.
‘Institutional capture’ 🏫🎯
"Oxford becomes first UK university to offer ChatGPT Edu to all staff and students" - imagining a lot of staff there pretty upset about, but also it says much about OpenAI's education strategy that it's targeted Oxford as its first institutional capture in the UK www.ox.ac.uk/news/2025-09...
Oxford becomes first UK university to offer ChatGPT Edu to all staff
OpenAI’s flagship GPT-5 model will be provided across the University and Oxford Colleges through ChatGPT Edu, a version of ChatGPT built for universities that includes enterprise-level security and
www.ox.ac.uk
September 19, 2025 at 5:41 PM
‘Institutional capture’ 🏫🎯
At IWCS soon, Matteo Radaelli shows 17 Norwegian LMs struggle to recover plausible covert events in complement coercion, in spite of positive (but not uniform across LMs) context effects 🤖💭 w/ Emmanuele Chersoni & Alessandro Lenci #compositionality #semantics
iwcs2025.github.io/program
iwcs2025.github.io/program
Conference Program
Düsseldorf
iwcs2025.github.io
September 19, 2025 at 9:27 AM
At IWCS soon, Matteo Radaelli shows 17 Norwegian LMs struggle to recover plausible covert events in complement coercion, in spite of positive (but not uniform across LMs) context effects 🤖💭 w/ Emmanuele Chersoni & Alessandro Lenci #compositionality #semantics
iwcs2025.github.io/program
iwcs2025.github.io/program
كتابي "علم اللغة العصبي" متاح الآن مجاناً باللغة العربية! شكراً للمترجمين الرائعين الذين جعلوا هذا ممكناً. أتمنى أن يصل إلى قراء جدد ويساهم في بحوث علم اللغة العصبي في العالم العربي.
#علم_اللغة_العصبي #العربية #ترجمة
#Neurolinguistics #Arabic #translation
www.hindawi.org/books/905090...
#علم_اللغة_العصبي #العربية #ترجمة
#Neurolinguistics #Arabic #translation
www.hindawi.org/books/905090...
اللغويات العصبية | جوزويه باجيو | مؤسسة هنداوي
www.hindawi.org
September 18, 2025 at 1:51 PM
كتابي "علم اللغة العصبي" متاح الآن مجاناً باللغة العربية! شكراً للمترجمين الرائعين الذين جعلوا هذا ممكناً. أتمنى أن يصل إلى قراء جدد ويساهم في بحوث علم اللغة العصبي في العالم العربي.
#علم_اللغة_العصبي #العربية #ترجمة
#Neurolinguistics #Arabic #translation
www.hindawi.org/books/905090...
#علم_اللغة_العصبي #العربية #ترجمة
#Neurolinguistics #Arabic #translation
www.hindawi.org/books/905090...
‘Machine language as somniloquy: How the neuroscience of sleep illuminates generative AI’ — New quick paper to appear in THINK, exploring the idea that sleep talk and AI-generated language unfold in cognitive conditions with similar restrictions 🧠🤖😴
lingbuzz.net/lingbuzz/009...
lingbuzz.net/lingbuzz/009...
Machine language as somniloquy: How the neuroscience of sleep illuminates generative AI - lingbuzz/009303
Can machines mean what they say? Generally not, because current AI systems lack the cognitive capacities that allow humans to recover meaning from and infuse meaning into language. However, from a cog...
lingbuzz.net
September 17, 2025 at 1:06 PM
‘Machine language as somniloquy: How the neuroscience of sleep illuminates generative AI’ — New quick paper to appear in THINK, exploring the idea that sleep talk and AI-generated language unfold in cognitive conditions with similar restrictions 🧠🤖😴
lingbuzz.net/lingbuzz/009...
lingbuzz.net/lingbuzz/009...
Reposted by Giosuè Baggio
Interested in doing a PhD with me and lacns.github.io? Or with any of the incredible fellows in the IMPRS School of Cognition www.maxplanckschools.org/cognition-en - apply before Dec 1st at cognition.maxplanckschools.org/en/application
Language and Computation in Neural Systems
We are an international group of scientists consisting of linguists, cognitive scientists, cognitive neuroscientists, computational neuroscientists, computational modellers, computational scientists, ...
lacns.github.io
September 3, 2025 at 3:15 PM
Interested in doing a PhD with me and lacns.github.io? Or with any of the incredible fellows in the IMPRS School of Cognition www.maxplanckschools.org/cognition-en - apply before Dec 1st at cognition.maxplanckschools.org/en/application