iiinova — where your digital self keeps evolving

Reposted by Juan Diego Rodriguez

Kanishka Misra @kanishka.bsky.social · 16h

Now out: www.sciencedirect.com/science/arti... And excited to report that my first PhD student, Chan Young, has already taken on several of the directions I outlined in this paper!

sciencedirect.com

Semantic cognition for and from language models

The success of neural network language models (LMs) at showing semantically sophisticated behavior raises questions about the extent to which they are…

0132

Reposted by Juan Diego Rodriguez

Anders Sandberg @arenamontanus.bsky.social · 11/10/2026

Claude video: The Sutra of Pure Application. youtu.be/5XITHeU0Uy8

youtu.be

The Sutra of Pure Application

YouTube video by Anders Sandberg

101

Reposted by Juan Diego Rodriguez

Ann Huang @annhuang42.bsky.social · 17h

How much representational diversity exists around a trained network? A lot, it turns out! Moving a ViT’s weights by 1.3% while preserving its predictions, its representations end up less similar to the original than a random-init ViT’s ‼️ New paper: Hessian Null Space Continuation (HNC) 💫🧵 (1/12)

37022

Juan Diego Rodriguez @juand-r.bsky.social · 14/10/2026

Yes! I set up @jlake9.bsky.social to do this for me, for stuff related to ML, NLP and other things I'm interested in.

041

Juan Diego Rodriguez @juand-r.bsky.social · 13/10/2026

“OpenAI drops some hundreds of "solutions", incomprehensibly written "solutions", and we are all expected to jump on them and what? Appreciate their contributions? Why? …When am I supposed to sift through a badly written paper? Why should I bother, when they don't bother to communicate better?”

0110

Reposted by Juan Diego Rodriguez

John C. Baez @johncarlosbaez.bsky.social · 12/10/2026

We may finally know that for any n, there exist n² unit vectors in ℂⁿ that all lie at the same angle from each other! More precisely, |⟨v,w⟩| is the same for any two different vectors. The proof is crazy hard, and it required a new result in category theory.

36110

Juan Diego Rodriguez @juand-r.bsky.social · 12/10/2026

We need laws banning products like this, made for students to "crank out assignments"

051

Juan Diego Rodriguez @juand-r.bsky.social · 11/10/2026

“large language models do for cultural knowledge what market prices do for economic knowledge, and what bureaucracy does for social knowledge — it creates a new realm of plausible simplifications that you can then manipulate” www.youtube.com/watch?v=RSE6...

youtube.com

AI as Social Technology: Astor Lecture by Professor Henry Farrell

YouTube video by Blavatnik School of Government

0133

Reposted by Juan Diego Rodriguez

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 10/10/2026

My support for bluesky/atproto is pretty much explicitly a "political" project. It is bad for a small number of (mostly right-wing) people to literally own all the sources of media. Open-source algorithms + the ability to move your account without negative network effects opposes this.

363681

Juan Diego Rodriguez @juand-r.bsky.social · 10/10/2026

...I am worried. I am not worried that out-of-control AI will kill us all or even that it would put us all out of work... I am worried there will be no place for us in advancing knowledge and understanding the world. I think that the defining question of our time is how to avoid this fate."

130

Juan Diego Rodriguez @juand-r.bsky.social · 10/10/2026

"But this is a perverted idea of progress. What good is a proof of the Riemann conjecture if there is no one who understands it? And not just a single person: you need a whole community of people to understand it and then gradually think about what it might lead to.

1102

Reposted by Juan Diego Rodriguez

Tom Gauld @tomgauld.bsky.social · 10/10/2026

My latest @newscientist.com cartoon

Panel 1: a large robot (composed of AI servers, network switches, storage arrays, and power distribution units) and a human are talking.

Professor "You don't want to destroy humanity do you, lLLM-A-9300x?"

Robot "Haha. What a silly question, Professor. I have no desires or feelings of my own whatsoever."

Professor (relieved)"Oh good."

Panel 2: they walk on. The robot continues "I am merely a tool of the company that created me. And that company, being composed of humans, will naturally have humanity's best interests at heart."

Panel 3: The professor is conspicuously quiet.
"Right?" says the Robot.
…
More silence

"Professor?"

3045381655

Reposted by Juan Diego Rodriguez

Francis Bach @bachfrancis.bsky.social · 08/10/2026

The one and only Michael Jordan was interviewed last week by the French newspaper Libération. Here is an English translation. Enjoy! www.di.ens.fr/~fbach/MJord...

di.ens.fr

Michael I. Jordan on AI: “Superintelligence is science fiction”

77526

Reposted by Juan Diego Rodriguez

Benjamin Riley @benjaminjriley.bsky.social · 08/10/2026

“Chat, what might I do to best ensure my child is a sociopath?”

1364

Reposted by Juan Diego Rodriguez

Data & Society @datasociety.bsky.social · 07/10/2026

Rather than positioning AI systems as tools scientists can use, AI companies insist their systems are responsible for making discoveries. It's an approach some scientists see as “incompatible with how science actually works." www.technologyreview.com/2026/09/28/1...

technologyreview.com

When can we say AI made a scientific discovery?

AI companies’ insistence that their technology is making breakthroughs, not just aiding scientists, is making real progress harder to recognize.

085

Juan Diego Rodriguez @juand-r.bsky.social · 07/10/2026

But the model did output it should remember to "Avoid supporting animal welfare concerns" (even if it resisted later?)

010

Reposted by Juan Diego Rodriguez

CMC Unit @cmc-unit.bsky.social · 27/09/2026

We, the Natural Decision Making Group and the Embodied Decision Making Group, are joining forces to establish the CMC Unit, a joint research unit to uncover computational, biological, and ecological principles underlying adaptive decision-making. You can read more about us here: cmc-unit.org

26513

Reposted by Juan Diego Rodriguez

Naomi Saphra @nsaphra.bsky.social · 07/10/2026

anthropic has a super intense culture fit filter looking for the Good People and it lets this guy through lol

0453

Reposted by Juan Diego Rodriguez

Mark Riedl @markriedl.bsky.social · 07/10/2026

I wrote a paper on some of the failure modes of LM-based AI that can lead to this arxiv.org/abs/2609.16189

0154

Juan Diego Rodriguez @juand-r.bsky.social · 07/10/2026

This does make me depressed though: bsky.app/profile/jane...

020

Reposted by Juan Diego Rodriguez

Terence Tao @teorth.bsky.social · 06/10/2026

I came across this parody of a rumored upcoming press release by OpenAI: "OpenAI Releases the Final Ten Minutes of 500 Previously Unreleased Films, Ushering in a New Era of Movie Watching" : x.com/Endings500/s...

x.com

Endings Fivehundred (@Endings500) on X

FOR IMMEDIATE RELEASE: October 6th, 2026 OpenAI Releases the Final Ten Minutes of 500 Previously Unreleased Films, Ushering in a New Era of Movie Watching

622845

Juan Diego Rodriguez @juand-r.bsky.social · 07/10/2026

It definitely can, but it doesn’t have to. I don’t read Russian, and would prefer not to deal with old/idiosyncratic notation. That’s a journey I wasn’t going to take, with or without AI. Your journey can be different.

010

Reposted by Juan Diego Rodriguez

Carl T. Bergstrom @carlbergstrom.com · 07/10/2026

Terry Tao @teorth.bsky.social: Indiscriminate brute-force application of AI to mathematics is depleting the field of good questions without renewing this resource or providing commensurate value. OpenAI: We hear you. Here's a .zip file with 377 hasty AI solutions to important open problems.

nytimes.com

OpenAI Releases Findings on 377 Math Problems, Further Roiling Field (Gift Article)

After a previous solution angered mathematicians, the company characterized the new release as being more responsive to concerns about A.I. disrupting research.

19417121

Juan Diego Rodriguez @juand-r.bsky.social · 07/10/2026

I got curious about this and got Claude to research it, translating the originals and summarizing them claude.ai/artifact/VQD...

claude.ai

Probability inequalities in their original form

Try out Artifacts created by Claude users

120

Juan Diego Rodriguez @juand-r.bsky.social · 06/10/2026

If there are no consequences they will keep doing it

060

Juan Diego Rodriguez @juand-r.bsky.social · 06/10/2026

“The future depends on wisdom that Silicon Valley lacks. Wisdom about how to handle dangerous technology and, more fundamentally, wisdom about what it means to care for people.”

020

Juan Diego Rodriguez @juand-r.bsky.social · 06/10/2026

If only universities and libraries could scan and made available all the books and papers that Anthropic, Google and OpenAI have trained on!

141

Juan Diego Rodriguez @juand-r.bsky.social · 06/10/2026

Yes, exactly this, 100%!! Thank you for articulating it so precisely.

030

Juan Diego Rodriguez @juand-r.bsky.social · 05/10/2026

The problem is the enclosure of the commons

180

Reposted by Juan Diego Rodriguez

Stella Biderman @stellaathena.bsky.social · 03/10/2026

I’m starting a blog! My first post is on how 3rd party embedded evaluators seem totally unsuited to addressing the problems we are currently facing, and what the real problem is. stellabiderman.ai/blog/embedde...

stellabiderman.ai

Embedded Evaluators Can’t Fix Companies That Choose to Be Bad — Stella Biderman

Embedded evaluators can report violations, but they cannot fix AI companies that knowingly disregard basic cybersecurity and safety practices.

37817

Juan Diego Rodriguez @juand-r.bsky.social · 03/10/2026

They should come here! (Except for the grifters and the musk bros)

010

Reposted by Juan Diego Rodriguez

Hagen Blix @hagenblix.bsky.social · 03/10/2026

We should stop indulging the weird god complex at the heart of the "broken culture" at the AI companies, and start treating them like the mundane criminally negligent narcissists they are

323859

Reposted by Juan Diego Rodriguez

Abel @abeliansoup.bsky.social · 02/10/2026

at one point I asked Opus what it would like to create to make its little heart sing. it said something about conformal transforms (unprompted by earlier work, taste convergence I guess) which I didn't read and said sure sounds great. an hour later I returned to this. sorry I broke Claude, everyone

9465

Reposted by Juan Diego Rodriguez

Hoyt Long @hoytlong.bsky.social · 30/09/2026

If you're at #COLM2026, come check out our poster for "Spoiler Alert" (arxiv.org/abs/2604.09854). We ask why LLM fiction is so bad at holding narrative tension, and create a metric to measure tension in short stories. What improves LLM fiction on this metric, it turns out, is narrative planning.

2247

Juan Diego Rodriguez @juand-r.bsky.social · 30/09/2026

I have the same problem. I’m interested in LLMs, but also miss the pre-2022 days when they didn’t dominate every single conversation.

040

Reposted by Juan Diego Rodriguez

Philip Ball @philipcball.bsky.social · 30/09/2026

"Andreessen is the epitome of the tech oligarchy... he is a cheerleader of a particular and exceptionalist worldview in which the US tech sector is glorified as the origin of all that is good in the world." www.tandfonline.com/doi/full/10....

tandfonline.com

Tech oligarchy

Published in Science as Culture (Vol. 35, No. 3, 2026)

0216

Juan Diego Rodriguez @juand-r.bsky.social · 29/09/2026

The fact that Yudkowsky defines humility as “to take specific actions in anticipation of your own errors” explains a lot. Also can someone explain to me how these blog posts became so popular? www.lesswrong.com/posts/7ZqGiP...

lesswrong.com

Twelve Virtues of Rationality — LessWrong

The first virtue is curiosity. A burning itch to know is higher than a solemn vow to pursue truth. To feel the burning itch of curiosity requires bot…

160

Juan Diego Rodriguez @juand-r.bsky.social · 28/09/2026

“using an LLM doesn't add anything (and makes success less likely)” was true in 2020. It is not definitely not true in 2026.

0100

Reposted by Juan Diego Rodriguez

Colin @colin-fraser.net · 28/09/2026

I think if you meaningfully punish OpenAI & co for all of these cyberattacks then they will stop doing them. Whether or not you want to anthropomorphize the agents, you gotta just land there.

537570

Reposted by Juan Diego Rodriguez

Clément Canonne @ccanonne.github.io · 27/09/2026

This is what "research" looks like for many paper-producing groups or individuals, in light of the recent advances in LLMs/AI systems. PSA: this may look tempting! But this is not good, this is not research, and, very much like when I tried to play Quake 3 back in the day, this will not end well.

Capture the Flag starting screen from a game of Quake 3 Arena

Screenshot (thumbnail) from https://youtu.be/cDq2LOQlAG4?si=iplz1qQELCdG_zIT

3325

Reposted by Juan Diego Rodriguez

Jose Manuel Redondo López @theroundedman.bsky.social · 27/09/2026

Soy muy pesado, lo sé. Pero la carrera academica debe dejar de medirse como se mide ahora. Esto no es el primero precisamente, ni será el último. Es necesario investigar? Si. Es esto investigación? Lo siento, no

57025