Alfredo González-Espinoza

Scientist | Librarian -
Complex Systems & Computational Methods for Interdisciplinary Research -

Research & Data @CMULibraries @CarnegieMellon
#JuliaLang

[Nature Computational Science @NatComputSci]

📢In our latest Comment for the 5-year anniversary Series, Natnatee Dokmai argues that, while privacy-preserving GWAS has strong technical benchmarks, they should connect formal guarantees to trust, risk, and governance. https://t.co/edl1oqKada

🔓https://t.co/rVJy0bvAAy

[Science Vs @sciencevs]

A U.S. FDA committee is meeting today & tomorrow to talk about changing the rules for substances including PEPTIDES. People claim these can heal our bodies & make us look great.

But does the science track?? We dig in 👇👇👇

https://t.co/e3vdnMJMtO

[Alfredo González-Espinoza @Spiralizing]

@YanboZhang3 Do you have something more shareable?

[Michael Levin @drmichaellevin]

New #preprint - @YanboZhang3

"Intelligence from Learnable Novelty"
https://t.co/PQz0lPckcL

What if we optimize Epiplexity (https://t.co/YYjWdlqKzn @m_finzi @andrewgwils ) instead of measuring it? We have derived a closed-form approximation of Epiplexity and discovered a deep connection between it and intelligence. This allows us to reinterpret Epiplexity as a form of learnable novelty, providing a brand-new understanding of what intelligence is. By maximizing Epiplexity across various systems, all of them exhibited interesting behaviors:

Cellular Automata: Maximizing Epiplexity directly generates complex soliton interactions similar to Rule 110.
Image Encoders: It automatically causes the encoding to cluster, successfully categorizing different handwritten digits without supervision.
Reinforcement Learning: Introducing Epiplexity improves the performance of PPO in sparse reward tasks.

We also explored the relationship between the theory of learnable novelty, the free energy principle, and novelty search. We hope this work helps us better understand the nature of intelligence and its origins.

[Steven Strogatz @stevenstrogatz]

How fast is the universe expanding? Two reasonable ways of estimating the expansion rate give puzzlingly different results. I discussed what this mystery might mean with Nobel Prize winning astrophysicist Adam Riess - join us! https://t.co/GYSPwDaH5E https://t.co/9PzRMznv0e

[Gro-Tsen @gro_tsen]

Somewhat unsurprisingly, this thread 🧵🔽 triggered a lot of people who self-identify with the “idiot armed with ChatGPT pro subscription”, who imagine math is a competition and who now feel they're getting their revenge on “gatekeeping” academics being now “butthurt”. •1/16

[Thomas Bloom @thomasfbloom]

A lot of people finding counterexamples with AI here, hyping up the headline with "open for 20 years" or "open for 60 years" etc.

Please stop - this doesn't necessarily mean it was hard or important, it can just mean it was not important enough for anyone to seriously work on.

[Christian Szegedy @ChrSzegedy]

It's a cool result, but for context:

GRAFFITI conjectures are essentially pre-AI slop: computer-generated statements from the nineties that could not be refuted easily by automated means.

Most of them were never worked on, as they tend to be quite isolated from everything else.

[Nathan Lambert @natolambert]

New podcast with @xeophon on all things open models. More on Kimi K3, Qwen 3.8, GLM-5.2, Xi's WAIC speech, distillation, the open-closed Gap, and what's next.

Chapters:
00:00 Welcome & context
04:38 Living with / using Kimi K3
08:53 GLM 5.2’s continued role
12:47 How are the Chinese models this good?
17:41 Data, environments, and a tour of the Chinese labs
19:47 Roundup of Chinese providers: Qwen, DeepSeek, MiniMax…
24:08 The US open-model ecosystem
30:25 Frontier vs. near-frontier, and the cybersecurity case against bans
34:58 Distillation and the Ben Thompson debate
44:12 Predictions and a frontier tier list
48:36 Wrap-up

Hoping to keep doing a few more of these on @interconnectsai. Crucial times in AI, we're working hard to share our expertise.

[Stanford HAI @StanfordHAI]

When does personalizing interventions in medicine, education, or job training actually improve outcomes? Stanford HAI Associate Director @EmmaBrunskill and @Stanford postdoc Zhaoqi Li built a statistical test to measure when it is actually worth it: https://t.co/20UenWZHjs

[Stephen Wolfram @stephen_wolfram]

Remember busy beavers? Now meet wily weasels: Turing machines that hide their bugs for as long as possible. (The wily weasel function WW says how wily a Turing machine of a given size can be.)

[Santa Fe Institute @sfiscience]

There’s a puzzling paradox in anthropology: The larger a population gets, the better protected it should be from random fluctuations in births, deaths, and environmental conditions that could cause it to collapse. And yet, the archaeological record is rife with examples of large civilizations that have collapsed, like the Roman and Egyptian empires.

New research from SFI External Professor Marcus Hamilton, an anthropologist at the University of Texas at San Antonio, points to an unexpected culprit: cooperation. 

https://t.co/taqR3rSWb4

[Alfredo González-Espinoza @Spiralizing]

@dyamins I am not really sure about the implementation (it's not my work), but to me sounds like an interesting idea (will probably give more insights to the results you are getting with the entropy values, e.g. how actual hard tasks are?)

[Alfredo González-Espinoza @Spiralizing]

Would be cool to see this approached with epiplexity.

[Daniel Yamins @dyamins]

New substack alert! We present the mathematical theory showing how Contravariance (probably) explains convergent evolution between AI models and brains. https://t.co/3Hla7rXQIO The zippering theorems put pretty strong guarantees on such convergence. 1/

[Carnegie Mellon University @CarnegieMellon]

Our AI Science Foundry will receive $20 million from @NSF as an initial node in a national network of connected laboratories to transform how scientific discoveries are made. https://t.co/GqeyKGZfNS

[Melanie Mitchell @MelMitchell1]

The ultimate tldr on the "rogue AI" hacking incident. 😅

But it would be good to know the prompts OpenAI gave the model....

[kache @yacineMTB]

@mkratsios47 I don't remember letting anthropic scrape my GitHub profile as well as all of my online writing!

[Yanbo Zhang @YanboZhang3]

Few concepts are invoked across as many disciplines as intelligence, and few are theorized in as many incompatible ways. To statistics and machine learning it is extreme compression of data; to the study of complex systems it is the emergence of universal computation; in the interaction of an agent with its environment it is open-ended adaptive behavior. Each appearance has its own literature and its own objective function, and the literatures rarely meet. Here we show that these appearances follow from a single principle: the pursuit of learnable novelty. 🧵

[Alfredo González-Espinoza @Spiralizing]

@aran_nayebi Is neuro embodied? (Is the limit of machines the physical one?)

[Alex Kontorovich @AlexKontorovich]

Terry posted a beautiful “digestion” of the Jacobian counterexample: https://t.co/KNXv47UaAm