# DeepMind-Like Gaming AI: Incredible Driving Skills! ## Метаданные - **Канал:** Two Minute Papers - **YouTube:** https://www.youtube.com/watch?v=PAjlXQBGK8U - **Дата:** 27.08.2023 - **Длительность:** 5:09 - **Просмотры:** 73,919 ## Описание ❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambdalabs.com/papers 📝 The paper "MAESTRO: Open-Ended Environment Design for Multi-Agent Reinforcement Learning" and "The Power Particle-In-Cell Method" are available here: https://sites.google.com/view/maestro-ued https://www.seas.upenn.edu/~ziyinq/static/files/power.pdf https://dl.acm.org/doi/abs/10.1145/3528223.3530066 My latest paper on simulations that look almost like reality is available for free here: https://rdcu.be/cWPfD Or this is the orig. Nature Physics link with clickable citations: https://www.nature.com/articles/s41567-022-01788-5 🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible: Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Benji Rabhan, Bret Brizzee, Bryan Learn, B Shang, Christian Ahlin, Geronimo Moralez, Gordon Child, Jace O'Brien, Jack Lukic, John Le, Kenneth Davis, Klaus Busse, Kyle Davis, Lukas Biewald, Martin, Matthew Valle, Michael Albrecht, Michael Tedder, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Rajarshi Nigam, Ramsey Elbasheer, Richard Sundvall, Steef, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Timothy Sum Hon Mun, Torsten Reil, Tybie Fitzhugh, Ueli Gallizzi. If you wish to appear here or pick up other perks, click here: https://www.patreon.com/TwoMinutePapers Thumbnail background design: Felícia Zsolnai-Fehér - http://felicia.hu Thumbnail original prompt idea: LisaBaker Károly Zsolnai-Fehér's links: Twitter: https://twitter.com/twominutepapers Web: https://cg.tuwien.ac.at/~zsolnai/ ## Содержание ### [0:00](https://www.youtube.com/watch?v=PAjlXQBGK8U) Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. Today we are going to look at a new  AI agent that learned to play video   games really well and learned some…  really interesting things. I’ll leave   it at that for now and show it to you in a moment. Now, first things first, the goal was to create  a large number of procedurally generated levels, ### [0:21](https://www.youtube.com/watch?v=PAjlXQBGK8U&t=21s) Creating random levels these are created by a computer if you will. Here you see their AI driving on the Hungarian   Formula 1 track. The goal is that it would start  to play increasingly difficult levels and see if   it has proper zero-shot performance, that is,  can it play a level that it hasn’t seen yet. So   let’s switch to a previously unseen track. Great  job, little AI! But, why is this so interesting? ### [0:44](https://www.youtube.com/watch?v=PAjlXQBGK8U&t=44s) Unseen track It is interesting because of two things. One,  it learned some really interesting skills in   laser tag, a little shooter game. Look.   It learned to hide behind walls to evade   these shots. So what happens if a fight  breaks out? Well, not a problem. It can   also dodge these bullets. Or it gets better,  it can even run away from them. Loving it. And here comes the interesting part. Second,  this little AI learned to be a criminal. And   when you are asking, Doctor, how? Let me  show you these little shenanigans. It can   not only drive on new levels, but it has also  learned to force an opponent off of the road.    Ouch. It can also overtake by cutting corners.   Very creative. But the creative elimination of   the opponent does not stop there…oh my. Are you  seeing what I am seeing? Dear Fellow Scholars,   that is a hit and run. I can almost hear it say  “Bye-bye, see you next lap! ”. And if you happen ### [1:56](https://www.youtube.com/watch?v=PAjlXQBGK8U&t=116s) Hit and run to meet it again, this little criminal can also  block your cornering. And all this was trained on ### [2:06](https://www.youtube.com/watch?v=PAjlXQBGK8U&t=126s) Blocking corners a single graphics card. Not a cheap one, but one  graphics card nonetheless. Really cool experiment. So what is all this good for? Well, this  technique could be extended to games that   require more than 2 players and we could see what  kind of collaboration these AI agents can learn   together. Or… if you ask me, perhaps what kind of  criminal rings they create. Kidding, of course. And if you allow me, I would like to show you an  incredible computer graphics simulation paper.    This one addresses something that we call the  volumetric dissipation problem. What is that?    Well, look. We start the simulation with a  given number of particles, and over time,   due to the inaccuracies of our calculations,  they slowly disappear. Like a magic trick.    Don’t believe it? Well, look. We started with  this many particles, and now we only have this   many. This new technique is able to create these  magical simulations where the amount of volume   over time remains the same. So, does it also  apply to messier scenes than this? You bet your   papers it does! Check this out! The scene can  be an absolute mess, and it still works. And,   it not only works, but it can easily plugged into  already existing systems, and it will work up to   2. 5 times faster than previous methods. And  all this with volume preservation. Loving it. Now, one more interesting thing. Look. Only  500 people have seen this paper yet. So I am   really worried that if we don’t talk about  these incredible computer graphics research   works here on Two Minute Papers, perhaps no  one will talk about them. If you liked this,   please share it with your friends. Let’s keep  the spark alive and show these amazing graphics   papers to more new Fellow Scholars. Thanks for watching and for your generous  support, and I'll see you next time! --- *Источник: https://ekstraktznaniy.ru/video/13058*