Andrea Cossu retweeted
📢 Next at #CoLLAs2026: Learning from streams, evolving over time! 🌊🧬
Andrea Cossu (@Cossu94) showed us how diversity becomes a resource for Continual Learning. By combining streaming environments, complex physical systems, and neuroevolution, we can achieve true open-ended adaptation! 🧠⚡
🎙️ Huge thanks to Andrea for the visionary insights!
Since I was unable to find a slim Python implementation of the Avida #alife system, "I" built one in Python:
github.com/AndreaCossu/avida… (still testing it).
If you are aware of similar efforts or willing to contribute to this one, do let me know.
Reminder that the call for papers for the Streaming Continual Learning workshop at ECML 2026 expires on May 31st. We welcome a wide range of contributions (position, preliminary results, full papers).
Best of luck, and see you in Naples!
More info ➡️ecml26.scl-ai.org/
Andrea Cossu retweeted
「AIは訓練するのではなく、自ら”育つ”ようにすべきだ」
Sakana AIのリサーチャー、@sebastianrisi がポッドキャスト @EyeOn_AI に出演。進化的手法でニューラルネットワークを構築するNeuroevolution手法の概要、継続学習や人工生命(ALife)研究の今について語りました。
youtu.be/pPpDxB4N_mE
After a long tour, Streaming Continual Learning joins ECML.
There are still 2 months left to prepare your workshop paper, and a few more before coming to Naples to present it!
All the info you need is here:
ecmlpkdd26.scl-ai.org/
Feel free to DM me 📧 if you have any question
Had fun "writing" a simple Jax implementation of self-replicating neural networks (aka neural quine), following Gabor et al. (Artificial Life, 2022).
doi.org/10.1162/artl_a_00359
The code is here if you want to play with it👇
github.com/AndreaCossu/self-…
Andrea Cossu retweeted
Call for papers for 'Artificial Life for Science and Engineering'
We seek work applying ALife concepts and tools to model real-world systems and engineer solutions—and assist scientific discovery through open-ended and curiosity-driven search.
Call info: alifeforscience.github.io
Two new papers:
1. DERL learns physical systems by leveraging derivatives. It works surprisingly well in a continual learning setting. TMLR: openreview.net/forum?id=IbBC…
2. ComPhy is a modular approach that learns systems of PDEs via DERL alignment. ICLR 2026: openreview.net/forum?id=ER7z…
Andrea Cossu retweeted
In general, I think now is the time to bring back a lot of the ideas from artificial life and evolution into LLMs.
Back then, it was too hard to bootstrap everything.
Now we have strong LLM priors and we can get a lot more out of systems (ex. AlphaEvolve).
this is a glimpse into the 2030s: autonomous agents competing for compute and memory in real-time. human-written code will look like static museum pieces compared to these self-replicating, adversarial assembly warriors. most labs are still obsessed with benchmarks, while the real winners are building the infra for agents to fight and evolve in the wild. the red queen hypothesis applied to llms is the most underrated research direction right now.
Andrea Cossu retweeted
Smart (but not necessarily supersmart) robots that can learn to operate the tools & machines operated by humans can also build (and repair) more of their own kind: the "ultimate form of scaling" [1-5]. Life-like, self-replicating, self-improving hardware will change everything.
REFERENCES (easy to find on the web):
[1] Interview with J. Schmidhuber, Jazzyear, Shanghai, 2024. "Eventually, however, there will be entire societies of clever and not-so-clever physical machines that can collectively build from scratch all the things needed to make copies of themselves, mine the raw materials they need, repair broken robots and robot factories, and so on. Basically, a machine civilisation that can make copies of itself and then, of course, improve itself. Basically, I am talking about a new form of life, about self-replicating, self-maintaining, self-improving hardware, as opposed to the already existing, self-improving, machine-learning software."
[2] J.S. Baut den KI-gesteuerten Allzweckroboter! (Build the AI-controlled all-purpose robot!) F.A.Z., 2024
[3] J.S. TED talk at TED AI Vienna (2024): Why 2042 will be a big year for AI. About self-replicating, self-improving, life-like machinery.
[4] J.S. at Falling Walls: The Past, Present and Future of Artificial Intelligence. Scientific American, Observations, 2017. More on self-replicating, self-improving, life-like machinery conquering the universe within a few tens of billions of years.
[5] J.S. 1995-2025: The Decline of Germany & Japan vs US & China. Can All-Purpose Robots Fuel a Comeback? AI Blog, Jan 2025, based on [2].
Image taken from the cover of my 1987 diploma thesis on meta-learning, self-referential, self-improving systems
Andrea Cossu retweeted
i just got an art grant from the council for the arts at MIT!
*Tangible Dreams* will let visitors experiment and play with a physical neural network generating images real-time—by twisting knobs and switches, by reconnecting nodes together @ArtsatMIT
Andrea Cossu retweeted
I'm looking for several 𝗣𝗵𝗗𝘀, 𝗣𝗼𝘀𝘁-𝗗𝗼𝗰𝘀 and 𝗩𝗶𝘀𝗶𝘁𝗶𝗻𝗴 𝘀𝘁𝘂𝗱𝗲𝗻𝘁𝘀 to build my team at @Unipisa working on the exciting 𝘊𝘖𝘓𝘓𝘈𝘎𝘌 project starting in 𝗡𝗼𝘃𝗲𝗺𝗯𝗲𝗿 𝟮𝟬𝟮𝟱!
Check below all the details👇
📍lnkd.in/dcsjfetB
Andrea Cossu retweeted
#LOT Spring school
@biggiobattista presenting the lecture "Machine Learning Security: Lessons Learned and Future Challenges"
@Cossu94 @v_lomonaco @StefanoMelacci @aleitteb
Andrea Cossu retweeted
#LOT spring school
we continue our incredible lineup of speakers with Alexei Efros and the talk titled "Test-Time Training: Toward a Post-Dataset World",
@StefanoMelacci @Cossu94 @v_lomonaco @aleitteb
cai.diism.unisi.it/school/
Art, nature and Artificial Intelligence at LoT Spring school which just started.
With 50+ students, we are fully booked and eager to learn (over time).
@StefanoMelacci @v_lomonaco @TiezziMatteo Alessandro Betti, Marco Gori
And now for something completely different (but not entirely)...
finally arrived in Tübingen for the Guided Self-Organization conference: Machine Learning in Embodied Agents!
You never know where continual learning might pop up 🐥
Andrea Cossu retweeted
🚀 Learning Over Time (LOT) Spring School
📅 24-27 March 2025 | 📍 Siena, Italy
💡 Are you considering models that continuously adapt over time instead of learning "offline" from pre-designed-huge collections of data? 🚀
🔗sites.google.com/unisi.it/lo…
Tuscany's hills are worldwide famous 🌄
That's why we are organizing the "Learning Over Time" school inside the Certosa di Pontignano (Siena, March 2025)
The school is about machines that learn over time in dynamic environments, check it out!
cai.diism.unisi.it/school
The ESANN deadline has been extended to November 27!
You have an extra week to work on your continual/streaming learning paper before submitting to our special session 😁
esann.org/special-sessions#s…
See you in Bruges 🍺




















