DeepMind's AI Learned a Better Understanding of 3D Scenes

Backblaze:

📝 The paper "MONet: Unsupervised Scene Decomposition and Representation" is available here:

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
313V, Alex Haro, Andrew Melnychuk, Angelos Evripiotis, Anthony Vdovitchenko, Brian Gilman, Bruno Brito, Bryan Learn, Christian Ahlin, Christoph Jadanowski, Claudio Fernandes, Dennis Abts, Eric Haddad, Eric Martel, Evan Breznyik, Geronimo Moralez, James Watt, Javier Bustamante, John De Witt, Kaiesh Vohra, Kasia Hayden, Kjartan Olason, Levente Szabo, Lorin Atzberger, Marcin Dukaczewski, Marten Rauschenberg, Maurits van Mastrigt, Michael Albrecht, Michael Jensen, Nader Shakerin, Owen Campbell-Moore, Owen Skarpness, Raul Araújo da Silva, Richard Reis, Rob Rowe, Robin Graham, Ryan Monsurate, Shawn Azman, Steef, Steve Messina, Sunil Kim, Thomas Krcmar, Torsten Reil, Zach Boldyga.

Károly Zsolnai-Fehér's links:

[BALAGE]

I'd love to see how this AI reacts to images like Escher works, that have impossible geometry.

[mullgubben]

This is the only channel where I actually sit through the sponsor part at the end of the video. Just so I can hear you say "Thanks for watching and your generous support, and I'll see you next time!". Love it! Keep up the great work!

[ParameterGrenze]

Few people know that the last 5 episodes of 2 minutes papers were actually produced by an generative adversarial neural network, trained on newest AI papers and previous episodes as input. The quality is still astonishing, what a time to be alive !

[googacct]

With so many scholars falling out of their chairs, maybe they should add a seatbelt.

[martiddy]

When Károly says to "hold onto your papers", you know is going to be crazy stuff!

[connor-shorten]

Self-Supervised Learning resembles Unsupervised Learning in the sense that it does not require class labels,
however the task is structured in a similar way to Supervised Learning tasks, Great video!

[onslaught]

I absolutely love your videos, and try to recommend them to everyone I know. But sadly people aren't as interested in AI research as they should be. Still, love your videos and enthusiasm. It's nice learning about the latest research from someone who loves this stuff as much as me. Keep them coming!

[LordAlacorn]

General vision system is close - what a time to be alive!

[sumitar]

Please do a video on Google Deepmind Protein Folding

[lexer_]

The last time I watched a video from this channel was quite some time ago and now I am really impressed with how much your accent has improved. It bothered me quite a bit back then but now it really is not an issue at all at least for me.

[shmigelsky]

Wow, that's even more impressive than usual. It would be super exciting to show the network structure Deepmind came up with for the attention based variational autoencoder. Size, structure, parameter count and # of training examples used.

[ZardoDhieldor]

Unsupervised learning is always the most impressive!

[cyboticIndustries]

We need merch!.. Tshirt with "What a great time to be alive!" ("see you next time!" on the back) ... 'hold on to your hats' laptop stickers ...

[cmilkau]

I would love to hear a tiny bit more about the methods, even at the cost of hearing less about the results.

[pretzelboi]

Absolutely amazing. It's hard to believe we live in a time where this technology is possible.

[donaldviszneki]

Will become a supporter when more details are included. How about a rundown of the neural architecture, how it was trained, etc.?

[abhaypratap]

Is there any good source where I could get more information regarding detection of https malware

[tdk-in]

this algorithm will be useful when they make a fortnite AI and need to remove Ninja's smarmy, occluding face from their training data

[juhotuho]

this is so freaking fascinating, can't wait to see what more they come up with :D

[fireball]

Think of what this means for Waymo and autonomous driving too...