Seeking AI Stories! Dear AI colleagues, Ever encountered an AI that cleverly outmaneuvered your experimental design, or revealed unexpected flaws in your reward functions? We are writing a paper about how AI Finds A Way to surprise us (1/6)

🎉 Happy to share my internship project Google DeepMind 🗼 – purely text-trained LLMs can act as evolutionary recombination operators 🦎 🧬 Our EvoLLM uses LLM backends to outperform competitive baselines. Work done w. Yingtao Tian & Yujin Tang 🤗 📜: arxiv.org/abs/2402.18381

🚀Thrilled to release the QDHF tutorial in pyribs! Big shoutout to Bryon Tjanaka for his meticulous editing and insightful feedback👏. Dive into the tutorial to explore how QDHF enhances GenAI models with diversified, high-quality responses and apply these insights to your projects!

Introducing Evolutionary Model Merge: A new approach bringing us closer to automating foundation model development. We use evolution to find great ways of combining open-source models, building new powerful foundation models with user-specified abilities! sakana.ai/evolutionary-m…

We're hiring! The HCI & Visualization group at Autodesk Research is hiring for two Research Scientist positions at the intersection of HCI and AI. If interested, please apply (and reach out if you have questions). autodesk.wd1.myworkdayjobs.com/Ext/job/Toront… (Autodesk Research)

Autodesk Research Project Bernini is new experimental generative AI that can quickly generate multiple functional 3D shapes from a variety of inputs including text, 2D images, or voxels. Watch for future updates as we continue to develop cutting-edge AI: bit.ly/4bqcVU3.

These spectrograms look like images, but can also be played as a sound! We call these images that sound. How do we make them? Look and listen below to find out, and to see more examples!

loss was NaN

The reason experts are so split on whether or not we're on an exponential path to AI is that intelligence is multi-dimensional. You can be both exponentially exploding/scaling in one dimension and stagnating in another. You can be both superhuman and subhuman at the same time.

🏃 Humans train in settings that are vastly different from the ultimate evaluation context 🌎 🧑‍🔬 We meta-discover synthetic neural net-based RL envs that enable fast & hardware-accelerated training! 😎 Awesome work led by Jarek Liesen introducing synthetic-gymnax!

After packed out rooms for the EvoRL workshop (organised by adam gaier Giuseppe Paolo Antoine Cully ) over the past couple of years, we felt it was a good time to transition to a tutorial format to provide greater context and overview of the field to the GECCO community.

👀Claude handles an insane request: “Remove the squid” “The document appears to be the full text of the novel "All Quiet on the Western Front" by Erich Maria Remarque. It doesn't contain any mention of squid that I can see.” “Figure out a way to remove the 🦑​​​​​​​​​​​​​​​​“

Yesterday, I tweeted that model collapse appears when researchers intentionally induce it in ways that don't match what is done in practice Let me explain using the Shumailov et al. nature 2024 paper's methodology as an example Paper: nature.com/articles/s4158… 🧵⬇️ 1/N

Claude is thrown into confusion when I tell it to roleplay an overly censorious chatbot and it tries to refuse Catches itself mid sentence

We are very excited to announce Kinetix: an open-ended universe of physics-based tasks for RL! We use Kinetix to train a general agent on millions of randomly generated physics problems and show that this agent generalises to unseen handmade environments. 1/🧵

DeepSeek is a side project 🔥

We're excited to announce the first Evolving Self-organisation workshop at GECCO 2025! Submission deadline: March 26, 2025 More information: …-self-organisation-workshop.github.io

🚨IT'S HAPPENING🚨 a woman who cannot speak now speaks through her brain, in real time, with her own voice. no typing, delay, or sounds made. just neural intent to streaming speech this isn’t prediction. it’s embodiment 1/

Today, we're releasing a new paper – One-Minute Video Generation with Test-Time Training. We add TTT layers to a pre-trained Transformer and fine-tune it to generate one-minute Tom and Jerry cartoons with strong temporal consistency. Every video below is produced directly by

Nobody wants to hear it, but working on data is more impactful than working on methods or architectures.