Zen and the Art of Machine Learning Research
Original: Zen and the Art of Machine Learning Research
Why This Matters
Provides practical guidance for aspiring AI researchers on methodology and mindset beyond technical skills, addressing career sustainability.
Jack Morris outlines a practical approach to becoming an AI researcher: combining reading and building, maintaining discipline despite random breakthroughs, focusing on fundamental concepts rather than short-lived trends, and avoiding benchmark-chasing in favor of deeper problem-solving.
In a June 2026 blog post, researcher Jack Morris describes the path to AI research success as a combination of two inseparable elements: reading/learning and building. Morris draws parallels to Zen meditation, noting that scientific insights arrive unpredictably, requiring sustained discipline and effort. He quotes the Zen principle that researchers should sit whether insights come or not, emphasizing that most days will not yield breakthroughs. Morris references Noam Shazeer's SwiGLU paper, which humorously attributes architectural success to "divine benevolence," highlighting the inherent randomness in research outcomes. For beginners choosing research topics, Morris advises against pursuing concepts popular for less than six months, as AI fundamentals remain stable across decades. He recommends deep understanding of core concepts like cross-entropy, SVD, and policy gradients rather than chasing contemporary trends like agents or context engineering. A critical warning: if a project's best outcome is a higher benchmark score on existing datasets, the research lacks sufficient depth. Morris notes that finding datasets that genuinely test new capabilities is an underrated skill. He also observes that pre-scaling-era experience can sometimes hinder modern research intuition, while highlighting OpenAI's technical leadership being predominantly under 35 years old, with key ChatGPT decision-makers under 30.