Research Retrospectives: An interview with David Ha, Co-founder and CEO of Sakana AI

Interactive web-based visualizations serve as a vital medium for communicating machine learning research, offering deeper intuition than static PDFs. David Ha, a research scientist at Google, advocates for this approach, drawing on his background in finance and control systems to prioritize neuroevolution and complex systems over conventional gradient-based methods. By developing tools like EvoJax, he aims to align evolutionary strategies with modern hardware accelerators, facilitating cross-community collaboration. This shift toward hybrid publication models—combining traditional academic papers with interactive online demos—transforms how scientific knowledge is disseminated and defended. Observing natural systems and his own children’s exploratory behaviors further informs his research vision, highlighting the value of novel problem formulation and creative, non-rigid solutions in advancing the field of artificial intelligence.