WIP video about Markov Chains — Kindly Requesting Feedback!
Added 2025-07-21 16:48:14 +0000 UTCWe’d love to hear your feedback and suggestions as we wrap up work on our next video. Please let us know your thoughts!
Thank you so much for your time and help.
-Team Ve
Comments
Great idea for the video. My suggestion would be to touch the AI themes a bit more. For example self-supervised learning is a great example of Markov chains. How to teach drone to fly and avoid obstacles by itself? You could maybe think about splitting the video into multiple parts, as this is a very rich field to cover. Great work you guys ❤️
Nikolina
2025-08-02 18:09:23 +0000 UTCGreat presentation. You explore many of the questions that I have had over the years. Way back in the 80’s I had a small company that preceded google and yahoo and we had to develop a search engine for hospital equipment problems. I just didn’t think big enough and Yahoo and Google kicked our butts.
Bruce
2025-07-28 16:37:15 +0000 UTCAround 11:50, the nuclei in the background are moving the wrong direction. If the Neutron is traveling from left to right, the background should go from right to left
Grayson Mobley
2025-07-28 02:50:39 +0000 UTCGreat video. At 13:44 - are the prob matched with the 3 cases - traveling /leave - absorb /fission? At 14:46 - very nice discussion on the argument why Monte Carlo name was used Any mention of memory relative to Markov property? I know it comes at 25:25 but I wonder if it should not come earlier? You mention finance at the end, but maybe that should come earlier - and the memory property would tie in nicely there? Nice video as always with Veritasium.
Emmanuel Haven
2025-07-28 01:46:48 +0000 UTCWoah cliffhanger there Casper!!! :P I don't know if it'll fit, but I feel like the AI part doesn't get enough screentime? Though maybe it's worth saving it for a part 2, since the multiple-part videos tend to do pretty well... Like, essentially all of AI is just more and more sophisticated Markov chains to a certain degree (don't yell at me CS people I'm a physicist...), and for example when ChatGPT "first came out" I remember my CS friends saying "Honestly there hasn't been any major improvement in this version, this has existed for a while, I don't know why this is the one the public picks up on." People are using AI as a catch all term these days and I do feel like it'll help if we separated out the concepts
Origami Alice Zhang
2025-07-25 16:53:44 +0000 UTCThe YouTube channel Micro asserts that private equity now owns much or all of Veritassium, and this is why Team Ve is phasing out Derek Muller. Comments? https://youtu.be/hJ-rRXWhElI
Terry Bollinger
2025-07-24 06:34:25 +0000 UTCDigging the character intros on Nekrasov and Markov. Can we get a brief commentary on what Nekrasov and Markov thought that "free will" meant? That will help set the stage. This isn't the point of the video, but the first ~7min are spent on these two and a debate over probability and free will without mentioning (a) what each of them thought that free will meant and (b) how that relates to the mathematics of probability. Without knowing that, some of the commentary doesn't make much sense to me. @4:43 I was reading the letter and being surprised at how he isn't mincing words. Of course, I was also surprised by your reference to Bruns whom I consider a negligible quantity." "The unique service of P. A. Nekrasov, in my opinion, is namely this: he brings out sharply his delusion..." Harsh. Appreciating the visual of the letter and the larger context that it brings to Markov's character. @6:06 Curiosity: Would it impact the lesson to show Russian letters instead? Animation shows English letters, but the scientists were Russian. Pedantics like myself might wonder if it would look different with Russian lettering and their vowels and consonants, but the point is taken as it currently stands. @9:12/47 Audio: Sounds like "...the gat-jet...".
chromicacid
2025-07-23 05:07:39 +0000 UTCWould have liked names of people to remain longer. Introduce people to understand who they are. ENDING AT HOW MANY SHUFFLES ARE NEEDED--IS CRUEL.
robert kleps
2025-07-22 21:05:13 +0000 UTCVideo and audio fall out of sync for producer
robert kleps
2025-07-22 21:02:56 +0000 UTCThis could be one of the most important videos you have ever done. Great insight into AI. I wish I had some relevant input.
Edwin Pole II
2025-07-22 14:53:23 +0000 UTCReally cool video! It feels like there are so many cool examples to use. An important use case for an MCMC routine is in model fitting and error estimation in science. For examples the parameters of a light curve in Astronomy. I understand Netflix and Spotify use these techniques to come up with "recommendations" and 23 & Me use similar models to estimate your ancestry. There is so much to covere here, you could follow up with a whole video on just the Metropolis-Hastings algorithm
Phil
2025-07-22 14:05:34 +0000 UTCStarting at around 13:33 in the segment discussing traveling neutrons, the probability labels on the transition arrows of the directed graph do not match the accompanying narrative.
Harper Wilson
2025-07-22 01:29:19 +0000 UTCI always love how you ground the maths / science with history. And it was nice to see some more applications of Markov chains that arise by simplifying possibly complex situations into a choice of finitely many states. I would have loved a few more facts about how the simplicity of this "probabilistic finite automaton" allows also for simpler mathematical analysis (beyond Monte Carlo simulation). Here some production details I noticed. first minute,8:09, 8:19, 8:40, 8:59, 12:30, 14:30, 14:50, 15:45, 17:14, 17:28 18:22: lip sync problems 7:21 : the image shows a threshold of .33, but the explanation says ".66" 7:25 : the voice says "we get.21, so we move to a vowel". But the video shows moving to another consonant. 13:35 until 13:50: differing assignment of numbers to various edges betweeen graphics and speech The video seems cut short at the end after the question about shuffling.
Lionel Pöffel
2025-07-21 23:40:23 +0000 UTCAt 15 minutes into the video, Brian Hayes shows up. I did not know who he was. His webpage on the American Scientist references a 2013 article about Markov chains. That article is very congruent to much of the second half the video, which I think means Mr. Hayes deserves more credit. Granted, if not for this video, I would not have read the article today, or ever, so watching the video has been beneficial to my understanding of the topic.
Lawrence Judy
2025-07-21 21:22:24 +0000 UTCGreat video! Not sure if this is expected at this stage or if it is the patreon player, but there were quite a few sections where the audio was slightly out of sync with the video, I can rewatch and give specific time codes if needed
Jeffrey van Gogh
2025-07-21 21:19:38 +0000 UTC(1) Music volume seemed a bit too loud; it was interfering a bit with hearing some words. The music itself was great. (2) The video seemed very good overall; close to done with no obvious glitches that I could see. The phasing in of your new person is going well, with him carrying most of the load this time and doing a good job. (3) I liked that interview near the end, talking pointedly about the danger of assuming Markov chains with feedback still work okay (they do not -- an understatement!). That is very relevant to some impending, um, "disruptions" of our financial markets. Side comments: (a) Bringing in the human and social background issues up front is a nice touch and relevant in interesting ways to many math and physics topics. We are Markov chains of our upbringing and social contexts, yes? For example, once I discovered Hilbert's Calvinistic upbringing, many of his mathematical assumptions, especially regarding absolute perfection and infinite detail, fell into place with new clarity. (b) Markov chains seem to work well, at least in part, because they almost accidentally model causality in a universe with strict lightspeed limitations on the transfer of information. In such a universe, the local context must necessarily have a powerful influence on what comes next. Classical manifold and continuum maths, in contrast, often subtly sneak in assumptions that violate light speed, such as assuming that what you do locally by accelerating a spaceship creates a meaningful causality interpretation for the _entire_ universe -- the real source of block universe interpretations that undermine the reality of local dynamics. In contrast, Markov chains enable explicit recognition of such finite context limits, albeit without explicitly addressing their origins in physics.
Terry Bollinger
2025-07-21 20:46:54 +0000 UTCThe answer is 7 - I believe this will come at the end, possibly as part of the sponsor message. But maybe it should be better integrated into the video
Derek Muller
2025-07-21 19:59:58 +0000 UTCAh, Markov chains! They are a great topic and an immensely valuable technique. Long before Transformers, we got so many Markov chain robotics and AI research submissions at the US OSD DDR&E — the organization then above DARPA — that I told a co-worker I would be delighted if I never saw another Markov chain proposal in my inbox. :) I am glad to take a look, though. It is a topic with which I have some… um… familiarity.
Terry Bollinger
2025-07-21 18:48:09 +0000 UTCI love math that simplifies an otherwise impossibly complex concept.
Mike Mitchell
2025-07-21 18:34:14 +0000 UTCIs there more? I mean, the question remains, “So, how many shuffles [ by Ulam playing solitaire in his sickbed ] DOES it take to get a random arrangement of cards??” Hopefully, there’s or there’ll be a Veritasium episode delving into the idea of randomness. That said, I loved this program! Thank you very much for creating it. W. Sabatier
William Sabatier
2025-07-21 17:47:20 +0000 UTC
