Two Minute Papers (YouTube channel)

What the channel covers

Two Minute Papers takes a recent research paper and shows, in a few minutes, what it can do. The host, Dr. Károly Zsolnai-Fehér, has a background in light transport and rendering, so alongside AI the channel covers fluid, cloth and hair simulation, ray tracing and graphics. His trademark line, "What a time to be alive!", is also the channel's whole description.

The videos focus on results and demos rather than equations. That makes them a quick way to see what is now possible, not a way to learn how to build it.

Playlists worth browsing

• AI and Deep Learning - Two Minute Papers: the main collection of AI paper summaries.
• ChatGPT, GPT4, OpenAI, Stable Diffusion and more!: language and image models.
• DeepMind explained - AlphaFold and more!: work from DeepMind.
• NVIDIA RTX, AI, and more: graphics and AI research from NVIDIA.
• Fluid, Cloth and Hair Simulations (Two Minute Papers) and Light Transport, Ray Tracing and Global Illumination (Two Minute Papers): the simulation and graphics side.
• TU Wien Rendering / Ray Tracing Course: full university lectures on rendering, a different and much deeper format than the short videos.
• Two Minute Papers interviews: conversations with researchers.

Who it suits

Anyone curious about AI and graphics research: first-year students choosing a direction, developers who want to keep up without reading papers every week, and final-year students looking for an FYP idea grounded in recent work. Game developers and graphics programmers will find the simulation videos especially relevant.

How to learn from it without getting stuck in tutorial loops

This channel is for awareness, not skills. Use it on purpose:

• Treat it as a news feed. A video or two a week is plenty.
• Follow up on one paper a month. When something interests you, open the linked paper, read the abstract and figures, and see whether code is available. That is how a two-minute video turns into learning.
• Keep a list of ideas for projects, theses or FYPs, with the paper link next to each.
• Do not confuse demos with readiness. A striking result in a paper may be years from a product. Ask what data and compute it needed.

Pair it with on Sefism

• To actually build what you see, follow the Machine learning roadmap or the AI engineering roadmap.
• For graphics and simulation, the Game development roadmap and cheat sheet are the practical path.
• For intuition about how neural networks work, watch the 3Blue1Brown Neural Networks Series.
• Use Google Scholar to find the papers behind each video and what cites them.

Best for: a light, regular look at what is new in AI and graphics research.