Research
「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（ALif…
Sakana AI researcher Sebastian Risi appeared on the *Eye on AI* podcast to discuss neuroevolution — the approach of using evolutionary algorithms instead of gradient descent or reinforcement learn...
Sakana AI researcher Sebastian Risi appeared on the Eye on AI podcast to discuss neuroevolution — the approach of using evolutionary algorithms instead of gradient descent or reinforcement learning to optimize neural networks, taking inspiration from how nature evolved intelligence. He argues that traditional AI systems are limited by fixed architectures and one-time training, whereas evolutionary methods can create systems that continuously learn, self-organize, and grow their own neural structures over time. Risi also highlights Sakana AI's direction of combining evolutionary algorithms with large language models — using LLMs as mutation operators to navigate solution spaces more creatively, applying open-ended setups and quality-diversity techniques to avoid local optima.
Related
- Great to see this on X. This is something I created during my PhD time 20 years ago.
- EvoFlows: Evolutionary Edit-Based Flow-Matching for Protein Engineering
- Stochastic Gradient Descent in the Saddle-to-Saddle Regime of Deep Linear Networks
- A 'Neural Computer' is built by adapting video generation architectures to train a World Model of an actual computer that can directly simul…
Source: research