Fernando Rosas 🦋
@_fernando_rosas
Followers
5K
Following
13K
Media
174
Statuses
7K
All posts on the blue sky | Emergence, synergy, brains, AI safety, and mental health
London, England
Joined October 2017
Preprint time:.“Software in the natural world: A computational approach to emergence in complex multi-level systems”. A rigorous framework that not only identifies *when* emergence takes place, but explains *how* this happens computationally.
arxiv.org
Understanding the functional architecture of complex systems is crucial to illuminate their inner workings and enable effective methods for their prediction and control. Recent advances have...
19
91
370
I wonder if all these could be explained simply as an application of Koopman theory (ie that things just become linear if one adds enough dimensions…)?.
We really have not made a lot of progress on explaining the deep mystery of LLMs:. How does a model using matrix multiplication to predict the next word manage to simulate human thought well enough to do all the very human-like things it does? And what does that mean about us?.
1
0
6
Thread below 👇🏽.
Preprint time:.“Characterising high-order interdependence via entropic conjugation”. How a “conjugation” of information charts the space of possible entropic metrics of high-order phenomena and clarify the relationship between existent quantities.
0
0
1
Finally published:.“Characterising high-order interdependence via entropic conjugation”. The final version includes findings on similarities in the high-order profiles of spin systems and early music supporting a new characterisation of the O-information!.
1
5
17
This looks awesome! 👇🏽👇🏽👇🏽.
The Consciousness Issue is live!. This special issue explores the field's biggest questions with essays and interviews from experts like Anil Seth, Annaka Harris, Erik Hoel, and Peter Godfrey-Smith. Some highlights:. 🧵 (1/6)
0
0
15
Response to @anilkseth great piece “Conscious artificial intelligence and biological naturalism”. Thanks to the great team lead by the awesome @rhizoqualia!.
cambridge.org
Conscious artificial intelligence and biological naturalism
1
0
12
Preprint time:.“Compatibilist emergence for the science of consciousness”. How a multifaceted understanding of emergence can illuminate the link between consciousness and life!
4
34
154
🤩.
Ellie Pavlick asking the big questions at 9:30am: "What came first, the sum or the parts? Emergent compositionality in neural networks".📺 Watch live:
0
0
2
RT @UndercodeNews: Pioneer of Modern #AI Honored: Shun’ichi Amari Awarded the Kyoto Prize.
undercodenews.com
In the world of artificial intelligence, few names carry the weight and legacy of Shun’ichi Amari. A visionary researcher whose work laid the mathematical
0
3
0
Open postdoc position at SFI!! 👇🏽👇🏽👇🏽.
Collaborate with researchers from around the world. Chart your own transdisciplinary path. Applications for the 2026 Santa Fe Institute Complexity Postdoctoral Fellowships are open!. With competitive salary, collaborative and discretionary funds, a generous benefits package
0
1
5
RT @MacrinePhD: Inspired by the 'brain in a vat' experiment, new research explores "AI in a vat" for safe agent sandboxing. 🧠 .@_fernando_r….
lesswrong.com
Drawing inspiration from the ‘brain in a vat’ thought experiment, this blogpost investigates methods to simplify world models that remain agnostic to…
0
2
0
And if you don’t think world models are a big deal, please check yesterday’s release of DeepMind of Genie 3:. … and this fantastic coverage by @MLStreetTalk :.
deepmind.google
Today we are announcing Genie 3, a general purpose world model that can generate an unprecedented diversity of interactive environments. Given a text prompt, Genie 3 can generate dynamic worlds...
0
2
4
Original paper + thread 👇🏽.
Our manuscript “AI in a vat: Fundamental limits of efficient world modelling for agent sandboxing and interpretability” . … has been accepted for the RL Conference!. 🧵👇🏽.
1
0
1
New bogpost about our paper on world models for safe and interpretable AI sandboxing:. Including funky ideas such as what makes a world model, the metaphysical and empirical components of Bayesian beliefs, and the mathematics of the Kantian noumenon….
lesswrong.com
Drawing inspiration from the ‘brain in a vat’ thought experiment, this blogpost investigates methods to simplify world models that remain agnostic to…
1
2
7
Awesome conversation about intelligence and stupidity 👇🏽.
SFI President David Krakauer discusses the distinction between knowledge and intelligence in the age of AI, the concept of “exbodiment,” and what true emergence might mean for large language models. Watch the full interview with Machine Learning Street Talk:
0
1
8
RT @_TheTransmitter: 𝐖𝐡𝐚𝐭 𝐰𝐞 𝐚𝐫𝐞 𝐫𝐞𝐚𝐝𝐢𝐧𝐠 “Top-down and bottom-up neuroscience: overcoming the clash of research cultures,” by @_fernando_ro….
nature.com
Nature Reviews Neuroscience - As scientists, we want solid answers, but we also want to answer questions that matter. Yet, the brain’s complexity forces trade-offs between these desiderata,...
0
4
0
Hot take:.Any system trained to optimise a single function (irrespective of what that function encodes) ends up resembling someone who only cares about one single thing — a psychopath.
4
2
22
RT @geoffreyirving: I am very excited that AISI is announcing over £15M in funding for AI alignment and control, in partnership with other….
0
27
0
Great news article from the Basque Center for Applied Mathematics about our recent work:.
Finally published:.“Explosive neural networks via higher-order interactions in curved statistical manifolds”. Enhancing the capabilities of recurrent neural networks by deforming their geometry!.
0
2
12
RT @farairesearch: 📺 More Singapore Alignment Workshop talks are live! .Learn from @furongh @tzhang1989 @jiaming_pku @Animesh43061078 @shi_….
0
24
0