An interview article with Yutaro Yamada of Sakana AI, one of the lead authors of the AI Scientist paper recently published in @Nature magazine, has been published in Nikkei Digital Governance. nikkei.com/prime/digital-gov… In this article, Mr. Yamada provides a comprehensive explanation of the current state of AI-driven scientific research, covering both the achievements of AI Scientist and its challenges. As mentioned in the article, one of the important findings from this research is that we experimentally demonstrated that the quality of generated papers can improve as the performance of the underlying AI model improves. "Being published in a top-tier scientific journal means that Sakana AI's paper-creation system has been recognized as having 'passed' the scientific version of the Turing test." The article also accurately addresses the current limitations of AI Scientist. What matters to us is not just what today's models can do. The Nature paper's achievement is that we quantitatively tracked the relationship between model performance and research quality, thereby revealing what could be called "scaling laws for scientific research." We believe this provides valuable insights for broader discussions about the future of science. "In the paper published in Nature this time, we demonstrated that if the performance of the AI model underlying the system improves, there is potential to enhance the quality of generated papers, and future developments are expected. Yamada of Sakana AI states: 'I want to improve the quality of automated research and help human researchers achieve surprising results.'" The article also mentions that AI-generated papers are watermarked and that the experiments were conducted under the approval of ethics committees and academic societies. In an era when AI is significantly transforming the nature of science, Sakana AI will continue to explore the possibilities of AI in science through ongoing dialogue with the scientific community. Blog: sakana.ai/ai-scientist-natur… [Translated from EN to English]
@hardmaru
-
Can Codes and Symbols Emerge in Neural Networks?
By
–
Can codes and symbols emerge in a single neural network? Thinking from the "Neural Computers" paper @rmaruy
rmaruy.hatenablog.com/entry/… [Translated from EN to English] -
Training Session with Defense Research Institute: AI Technology and National Security Contribution
By
–
Recently, we had the pleasure of welcoming trainees from the National Institute for Defense Studies (NIDS) at our company! 🐟✨Sato and Ishii from our Defense and Intelligence Division took the stage to introduce the latest AI trends and our company's initiatives in the defense and intelligence sectors. There were many sharp questions from field experts, and we felt the weight of the responsibility.It was an extremely valuable time for us to think together about how AI can contribute to the nation's security going forward. [Translated from EN to English]
-
Neural Computers: AI Systems That Become Computing Environments
By
–
A "Neural Computer" is built by adapting video generation architectures to train a World Model of an actual computer that can directly simulate a computer interface. Instead of interacting with a real operating system, these models can take in user actions like keystrokes and mouse clicks alongside previous screen pixels to predict and generate the next video frames. Trained solely on recorded input and output traces, it successfully learned to render readable text and control a cursor, proving that a neural network can run as its own visual computing environment without a traditional operating system. arxiv.org/abs/2604.06425 Cool work by @MingchenZhuge @SchmidhuberAI et al.! Mingchen Zhuge (@MingchenZhuge) 🫱 Introducing 𝐍𝐞𝐮𝐫𝐚𝐥 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫s: 𝐰𝐡𝐚𝐭 𝐢𝐟 𝐀𝐈 𝐝𝐨𝐞𝐬 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐮𝐬𝐞 𝐜𝐨𝐦𝐩𝐮𝐭𝐞𝐫𝐬 𝐛𝐞𝐭𝐭𝐞𝐫, 𝐛𝐮𝐭 𝐛𝐞𝐠𝐢𝐧𝐬 𝐭𝐨 𝐛𝐞𝐜𝐨𝐦𝐞 𝐭𝐡𝐞 𝐫𝐮𝐧𝐧𝐢𝐧𝐠 𝐜𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐢𝐭𝐬𝐞𝐥𝐟? Beyond today's conventional computers, agents, and world models, Neural Computers (NCs) are new frontiers where computation, memory, and I/O move into a learned runtime state. We ask: whether parts of runtime can move inward into the learning system itself. This is our first step toward the Completely Neural Computer (CNC): a general-purpose neural computer with stable execution, explicit reprogramming, and durable capability reuse. Work done with Mingchen Zhuge (@MingchenZhuge), Changsheng Zhao, Haozhe Liu (@HaoZhe65347 ), Zijian Zhou (@ZijianZhou524 ), Shuming Liu (@shuming96 ), Wenyi Wang (@Wenyi_AI_Wang ), Ernie Chang (@erniecyc ), Gael Le Lan, Junjie Fei, Wenxuan Zhang, Zhipeng Cai (@cai_zhipeng ), Zechun Liu (@zechunliu ), Yunyang Xiong (@YoungXiong1 ), Yining Yang, Yuandong Tian (@tydsh ), Yangyang Shi, Vikas Chandra (@vikasc), Juergen Schmidhuber (@SchmidhuberAI) — https://nitter.net/MingchenZhuge/status/2042607353175097660#m
-
FunAI Lab Visits Sakana AI: Research Discussion and Presentations
By
–
Today we visited Japan's hottest AI startup @SakanaAILabs🎏🇯🇵! We met their research scientists and discussed the implications and impact of some their works like "The AI scientist" and "Continous Thought Machines". We presented our @FunAILab works, "Better Language Models Exhibit Higher Visual Alignment" and "Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning". Got lots of cool questions and discussions! Thanks, Masanori Suganuma, @_yutaroyamada, @ciaran_regan_
-
AI Should Be Nurtured, Not Trained: The Promise of Neuroevolution
By
–
"AI should not be trained, but should 'grow' on its own." Sakana AI researcher @sebastianrisi appeared on the podcast @EyeOn_AI. He discussed an overview of the Neuroevolution method, which constructs neural networks through evolutionary approaches, and talked about the current state of continual learning and artificial life (ALife) research. piped.video/pPpDxB4N_mE [Translated from EN to English]
-
Abstract Fish in Japan Generated in Pure SVG Code
By
–
"An Abstract Fish in Japan" (Pure SVG code generated by Sakana Chat) [Translated from EN to English]
-
Self-Driving Sushi: Autonomous Food Delivery Innovation
By
–
self-driving sushipic.twitter.com/v53GyS6MyB
— hardmaru (@hardmaru) 3 avril 2026self-driving sushi
-
Serial Experiments Lain 1998 – Machine Learning References
By
–
Serial Experiments Lain (1998) nitter.net/hardmaru/status/828047… hardmaru (@hardmaru) Machine Learning Enthusiast — https://nitter.net/hardmaru/status/828047097366667264#m [Translated from EN to English]