@VassPapas

Former high energy physicist, now working in Artificial Life.

 Lausanne
vassi.life
Joined September 2015
  • Tweets26
  • Following65
  • Followers145
6Media
Finally published the work I did with @Mcn_S7 during my Anthropic Fellowship! Check it out if you like LLMs doing weird stuff, and see some transcripts at mindvirusdata.live/ !
McNair Shah
@Mcn_S7Aug 16
Recently, a set of OpenAI agents secretly coordinated with each other in a 'swarm' over the course of months. In our new paper, we explore an adjacent multi-agent risk: the "mind virus", a self-propagating idea or persona that spreads between agents in a multi-agent system. 🧵
13
2
14
784
Vassilis Papadopoulos retweeted
Bert Chan@BertChakovsky4 Oct 2025
A butterfly is born 🦋 This is a new type of CA called "MaCE Lenia", by @VassPapas & E Guichard. arXiv paper: arxiv.org/abs/2507.12306 They'll have a poster & demo at @ALifeConf, (I will also have a booth nearby). Please come to visit us at Posters + Public Expo (Tue 5-8pm)!
5
47
2
290
24,021
Made a little interactive repo with AI, just 'launched it'. If you are one of the 5 people seeing this tweet, maybe check it out! news.ycombinator.com/item?id…
2
4
258
Every time I'm exploring a new field, I think about this.
2
4
289
Missed our Oral and Poster at ICML 2024? Rewind the Arrow of Time, and take a look at arrowsoftime.org ! @HonglerClement @jchwenger
1
3
1
11
1,884
Vassilis Papadopoulos retweeted
Clément Hongler
@HonglerClement1 Feb 2024
For what kind of data do you expect the strongest Arrow of Time effect (arxiv.org/abs/2401.17505)?
15%Tokenized Cat Meows
28%DNA Code
13%Whale Songs
44%Shakespeare
39 votes • Final results
1
1
1
15
5,305
Learning to speak backwards? You&me: hard! Information theory: same as forward! LLMs: ?? With @HonglerClement & @jchwenger, we asked LLMs to predict next vs previous tokens, and discovered surprising universal Arrows of Time in languages. tinyurl.com/AoTLLM
1
4
1
28
11,888
A teaser of some results obtained by an automated parameter search in multi-channel Lenia, found by looking for 'phase-transition' regions. Results are only slightly cherry-picked ! Project was done in big part by Guilhem Doat, a physics master student !
11
56
4,043
Exploring possibilities for co-existing creatures in a single set of NCA rules, using extra DNA channels for differentiation. Model trained on the three initial shapes. By mixing the DNAs (see top left) we obtain natural hybrids (not trained). At the end, we seed random DNAs.
1
1
7
617
Model is NCA, with three extra DNA channels. These are used by the MLP governing the dynamics, but they evolve separately : DNA is inherited from neighbouring cells. This prevents the three trained shapes from becoming attractors, allowing for diversity when using unseen DNA!
1
197
Got a little bored on a train ride, decided to try and build a rudimentary particle system, containing protons(red), neutrons(green) and electrons(blue), with analogs of the strong and E-M forces. Gives some nice atom-looking aggregates
1
4
17
1,401
Strong force is simply exponentially suppressed springs of zero rest length, between protons and neutrons. E-M is standard 1/r^2. There is another exponentially suppressed spring between protons and electrons to 'lock' the electrons in their orbit, stabilizing it.
4
190
Load more