Back to index
OtherOngoing YouTube series, individual episodes run a few minutes each·Free on YouTube

Two Minute Papers (Karoly Zsolnai-Feher)

4.0

The best way I know to stay excited about where AI research is heading, delivered in bite sized episodes that anyone can follow. Just be clear with yourself about what it is, which is a window onto the field rather than a course that teaches you to build anything.

What We Liked

  • Makes cutting edge research approachable without demanding you read the papers yourself
  • Consistently current, so you hear about important results not long after they land
  • Short, high energy episodes are easy to fit into any schedule and genuinely fun to watch
  • Links to the source papers, so you can dig deeper whenever something grabs you

What Could Be Better

  • It is a news and inspiration channel, not a course, so you will not learn to implement anything from it
  • The relentless enthusiasm and the catchphrases wear on some viewers over time
  • Coverage skews toward visually impressive graphics and generative results, so whole areas get little attention
  • Depth is deliberately shallow, so you get the what and the wow but rarely the how

Detailed review

Two Minute Papers has been running for years, and it does one thing extremely well, which is to take a fresh research result, usually in AI, computer graphics or physical simulation, and explain in a few minutes why it matters and what makes it impressive. Karoly Zsolnai-Feher, who came out of a research background at TU Wien, has a knack for finding the striking demonstration in a paper and framing it so a general audience can appreciate it, and the trademark upbeat delivery, catchphrases and all, has built a large and loyal following. If your goal is to stay aware of what is happening at the frontier without carving out hours to read papers, it is hard to beat, and it links each episode to the underlying work so you can go as deep as you like on the ones that catch your eye. Where I want to be honest is about what it is not.

This is not a course and does not pretend to be one, so you will finish an episode knowing that something remarkable exists without knowing how to build it or even, in many cases, how it really works under the hood. The depth is intentionally shallow, the pace is fast, and the selection leans toward the visually spectacular, which means generative imagery, rendering and simulation get far more airtime than, say, the less photogenic corners of the field. The enthusiasm that makes it accessible also grates on some people after a while, which is fair enough and comes down to taste. My take is that you should treat Two Minute Papers as the trailer, not the film.

It is a superb way to keep your motivation up and your awareness current, and I genuinely enjoy it for that, but pair it with real courses and hands on work when you want to move from admiring the research to understanding and reproducing it.

[ final ]

The verdict.

A brilliant supplement and a terrible standalone. Watch it to keep your finger on the pulse and to remember why you found this field interesting in the first place, then go elsewhere when you actually want to learn the mechanics. On that basis it earns its place in anyone's rotation.