distill.pub

What It Is

Distill is an online scientific journal focused on clear explanations of machine learning. The site explicitly states that it operated from 2016 to 2021 and is currently on indefinite hiatus. It is not a course platform in the traditional sense; instead, it turns papers, conceptual explanations, visual experiments, and peer-reviewed articles into web-native content, emphasizing “clear, dynamic and vivid” scientific communication.

Core Content and Format

Its content covers frontier topics in machine learning, including graph neural networks, neural network interpretability, visual analysis of reinforcement learning, Bayesian optimization, GANs, t-SNE, RNN memory, feature visualization, and adversarial examples. The teaching/learning format mainly consists of long-form English articles, animated graphics, interactive charts, and explorable model examples. Some articles are marked as peer-reviewed and provide DOIs; the site also has an ISSN. The lineup of instructors and authors is very strong, with contributors or committee members associated with OpenAI, Google, DeepMind, MIT, Microsoft Research, Université de Montréal, and others, making it suitable for serious study and research reference.

Pricing and Certificates

The text does not show any reading fees, subscription costs, or course purchase information, so it can be viewed as an open-access reading resource. The site mentions a past $10,000 prize intended to reward excellent work in communicating and distilling ideas, but this is not a learner-facing scholarship or course price. In terms of certification, no course completion certificate is mentioned; its value comes more from peer review, DOIs, and its academic citation attributes.

Pros and Cons

Its strengths are high-quality explanations and strong visualization, which help readers build intuition around complex algorithms—especially useful for making up for the opacity of papers and the limitations of static PDF presentation. The drawbacks are also clear: it has no systematic course pathway, assignments or quizzes, learning community, or instructor Q&A; the content is research-oriented and has a relatively high barrier for beginners; and since updates have been paused since 2021, it is less timely than continuously maintained courses or blogs.

Who It’s For and Access in China

It is better suited to graduate students with a foundation in machine learning, AI engineers, researchers, and advanced self-learners. It is also suitable as material for reading groups or as a reference for interactive instructional design. The teaching language is English, and the text does not provide a Chinese version. Access from mainland China cannot be determined based solely on the crawled text, so it is marked as unknown.