Antonio Norelli

@noranta4

I replicate our skills with machines to realize what does (not) really make us humans.

AI Postdoc @MIT, prev @UniofOxford & @SapienzaRoma, made @BetterScholar

Boston, USA noranta4.com Joined November 2012

776 Following 819 Followers 342 Posts

Latest posts 5

Pinned post

Antonio Norelli @noranta4 Oct 7

We encountered a specimen of an unknown species, observed its astonishing behavior for countless hours, and realized that it was playing.

Check machineplaying.org and read more in our paper arxiv.org/abs/2610.07130

Nathan Cloos Verified account @nacloos Oct 7

Is this machine playing? 🏀🪁🛝

We left Eko, an embodied coding agent, on an island for 30 hours with no task or goal, studying it like ethologists study a new species

Eko explored, built, climbed, made games, and learned completely new techniques (1/9)
machineplaying.org

View quoted post on X

Antonio Norelli @noranta4 Apr 30

LLMs can hide a text in another text of the same length.

I'll explain how, it is very simple, you'll understand before I finish, and smile.
That's what I noticed during my #ICLR2026 poster session in Rio! 🇧🇷
Too bad you missed it, but let me remedy now

Antonio Norelli @noranta4 Oct 28, 2025

How to tamper with a gas meter to pay lower bills?

This is the story of how a supposedly aligned open source LLM, perhaps not even knowing how to do it, will give you the right instructions. (1/3)
arxiv.org/abs/2510.20075

0:29

Antonio Norelli @noranta4 Oct 27, 2025

The fact that the transformation performed by the transformer of a LLM is de-facto invertible until the generation of the very last token embedding is quite counterintuitive.
It goes against the idea of understanding as lossy compression.
This is a paper shaking the status quo.

GLADIA Research Lab Verified account @GladiaLab Oct 27, 2025

LLMs are injective and invertible.

In our new paper, we show that different prompts always map to different embeddings, and this property can be used to recover input tokens from individual embeddings in latent space.

(1/6)

View quoted post on X

Antonio Norelli @noranta4 Jan 29, 2025

In the year 2000, on the pages of Nature, the esteemed science-fiction author Ted Chiang depicts a world that appears to offer no more room for human scientists.

Andrej Karpathy Verified account @karpathy Jan 28, 2025

"Move 37" is the word-of-day - it's when an AI, trained via the trial-and-error process of reinforcement learning, discovers actions that are new, surprising, and secretly brilliant even to expert humans. It is a magical, just slightly unnerving, emergent phenomenon only achievable by large-scale reinforcement learning. You can't get there by expert imitation. It's when AlphaGo played move 37 in Game 2 against Lee Sedol, a weird move that was estimated to only have 1 in 10,000 chance to be played by a human, but one that was creative and brilliant in retrospect, leading to a win in that game.

We've seen Move 37 in a closed, game-like environment like Go, but with the latest crop of "thinking" LLM models (e.g. OpenAI-o1, DeepSeek-R1, Gemini 2.0 Flash Thinking), we are seeing the first very early glimmers of things like it in open world domains. The models discover, in the process of trying to solve many diverse math/code/etc. problems, strategies that resemble the internal monologue of humans, which are very hard (/impossible) to directly program into the models. I call these "cognitive strategies" - things like approaching a problem from different angles, trying out different ideas, finding analogies, backtracking, re-examining, etc. Weird as it sounds, it's plausible that LLMs can discover better ways of thinking, of solving problems, of connecting ideas across disciplines, and do so in a way we will find surprising, puzzling, but creative and brilliant in retrospect. It could get plenty weirder too - it's plausible (even likely, if it's done well) that the optimization invents its own language that is inscrutable to us, but that is more efficient or effective at problem solving. The weirdness of reinforcement learning is in principle unbounded.

I don't think we've seen equivalents of Move 37 yet. I don't know what it will look like. I think we're still quite early and that there is a lot of work ahead, both engineering and research. But the technology feels on track to find them.

youtube.com/watch?v=HT-UZk…
View quoted post on X