GrĂ cies a @joanburdeus.bsky.social per la conversa. GrĂ cies per lâesforç dâacostar aquestes idees a la gent
@gemmadlc.bsky.socialGetting ready for my talk for the Natural Philosophy Forum on Friday with the help of Mastroianni đ
Gemma De les Coves
@gemmadlc.bsky.social
ICREA Research Professor in Barcelona
@gemmadlc.bsky.social@gemmadlc.bsky.socialGrĂ cies per aquesta oportunitat tan bonica de parlar d'algunes idees precioses... GrĂ cies, Joan, quin interlocutor!
@gemmadlc.bsky.social
@ara.catReposted by @gemmadlc.bsky.socialA WhatsApp hi tĂ© un fotograma de 8 1/2, de Fellini, perquĂš Ă©s una pel·lĂcula plena d'autoreferĂšncies: Mastroianni Ă©s al mateix temps personatge i director del film; ...
@gemmadlc.bsky.socialIt can be translated to English with the help of some artificial friend...
@gemmadlc.bsky.socialI've written an article for The Conversation about what I find most exciting these days: The reach of self-reference and negation in science, philosophy and beyond. Perhaps art?
@gemmadlc.bsky.socialNo, a priori it doesnât arise from such assumptions Have a look at our âAn invitation to universality in physics, computer science and beyondâ with TomĂĄĆĄ Gonda
@gemmadlc.bsky.social(*) assuming I can fly after yesterdayâs heavy rains
@gemmadlc.bsky.socialNext Friday Iâll give a talk (*) at the Natural Philosophy Forum at Johns Hopkins University.
Looking forward to this opportunity and to meeting Profs. Jenann Ismael and Sean Carroll (@seanmcarroll.bsky.social )!
Did I ever mention Iâd like to be a natural philosopher instead of a physicist? đ
@gemmadlc.bsky.social(*) assuming I can fly after last nightâs heavy rains
@gemmadlc.bsky.socialNext Tuesday I'll be in Yale talking about our adventures to understand the reach of universality in physics, machine learning and beyond.
Thanks to Prof. Jack Harris for hosting me. Looking forward to it!
@gemmadlc.bsky.socialI was willing to treat AI as an assistant. But if itâs a competitor, I donât know what to do. To be clear, the AI-generated papers offer little insight for a human like me, as they are sketchy in key conceptual parts, and very technical in others.
@gemmadlc.bsky.socialSome of the AI-generated papers in that project are dated March 2026, but we couldnât find them 3 weeks ago after an exhaustive literature search. The papers are just published in their website and cannot be traced elsewhere. They cite very generic sources.
@gemmadlc.bsky.socialNow we found an AI generated website with a so-called flagship program, which contains essentially all our ideas, our wording for concepts, our mistakesâŠ
@gemmadlc.bsky.socialSomething âcurious" happened to me. I'm working with a student on a project, who's been interacting with ChatGPT over the past 2 weeks to help him understand category theory concepts. Inevitably, he shared some ideas with ChatGPT, although he disabled the function to be trained on the dialogue.
@gemmadlc.bsky.socialWeâll kick off with a session on the uncomputability of halting and another on Lawvereâs fixed point theorem.
@gemmadlc.bsky.socialNext week weâll start a reading group on undecidability with Profs. Eva Miranda and Carles Casacuberta (mathematicians, category theory experts).
Itâs open to students - several good students already signed up.
@gemmadlc.bsky.socialThis form of universality for spin models is essentially the same as universality for neural networks.
So with this idea you can show how RBMs or DBMs, which only have 2-body interactions, can approximate (essentially) any probability distribution.
In case it helps...
@gemmadlc.bsky.socialThis may be related to our universality results, which show that two-body interactions suffice to generate higher body interactions.
We also provide sufficient conditions for the onset of this expressive power
@gemmadlc.bsky.socialTambĂ© entrevisten a MartĂ Dominguez sobre aquesta traducciĂł a lâĂșltim nĂșmero del SĂ piens - ben interessant!