❤️ Check out Lambda here and sign up for their GPU Cloud: https://lambda.ai/papers
Guide for using DeepSeek on Lambda:
https://docs.lambdalabs.com/education/large-language-models/deepseek-r1-ollama/?utm_source=two-minute-papers&utm_campaign=relevant-videos&utm_medium=video
Our neural rendering paper: https://users.cg.tuwien.ac.at/zsolnai/gfx/gaussian-material-synthesis/
Sources:
https://x.com/MattJedrzejewsk/status/1906835418588590318
https://x.com/MattJedrzejewsk/status/1907145560056078581
https://x.com/egeberkina/status/1906089394219491782/photo/1
https://x.com/cocktailpeanut/status/1906983829035974890
https://x.com/cocktailpeanut/status/1906956398136811820
https://x.com/ImSamThompson/status/1906860220703486227
https://x.com/Salmaaboukarr/status/1904868239278010389
https://x.com/reliantlane/status/1905674157293273333
https://x.com/Salmaaboukarr/status/1906996179872669811
https://x.com/ClaireSilver12/status/1906870887669801453
https://x.com/TheJasonRink/status/1906902317552586998
https://x.com/alokepillai/status/1905370346758234435
https://x.com/Gardraw/status/1907359888147157357
https://x.com/umesh_ai/status/1906257368515858748
📝 My 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:
Benji Rabhan, B Shang, Christian Ahlin, Gordon Child, John Le, Juan Benet, Kyle Davis, Loyal Alchemist, Lukas Biewald, Michael Tedder, Owen Skarpness, Richard Sundvall, Steef, Taras Bobrovytsky, Thomas Krcmar, Tybie Fitzhugh, Ueli GallizziIf you wish to appear here or pick up other perks, click here: https://www.patreon.com/TwoMinutePapers
My research: https://cg.tuwien.ac.at/~zsolnai/
X/Twitter: https://twitter.com/twominutepapers
Thumbnail design: Felícia Zsolnai-Fehér - http://felicia.hu
Оглавление (2 сегментов)
Segment 1 (00:00 - 05:00)
Yes, we know the new ChatGPT is good at making images, but man, after a few more days of experimenting, this is so much better than we previously thought. Check out these 8 amazing examples that you can also run right now for free. One, yes, you can generate images with it. But no one said that they have to stay as images. Give it to another AI and make a 3D object out of it. Yup, then you can rotate them around, change the lighting and look at those beautiful specular reflections. Loving it. This is the Second Law of Papers. Do not look at these works as islands. Everything is connected. And don’t forget, you can even plug it into a text to video system as well. Runway’s GEN4 just appeared and it is amazing at that. Their previous version is less than a year old, and look at that improvement. All this in less than a year. Wow. Two, it can make cooking recipes come alive, I really like how this is like a flowchart that shows what you should be doing when, and what you should get as a result. Perfect! But, wait a second. If it knows the result of the mixing, can we actually run some experiments with it without actually doing the experiment? Let me explain. First, mix a cup of olive oil with water. Then, add some boba to it. Now, some red food syrup. It looks kinda great, so let’s ruin it now. Yes! Milk! That would do it. Hmm…I expected worse. Now can we use ChatGPT as a time machine and let it sit for a month? See what happens then? It knows chemistry after all, I hope all those benchmarks were not for nothing. Let’s see… Oof! Oh my… I choose to believe this one. Also, using an image AI as a time machine. A fantastic idea. What a time to be alive! Three, it can teach you to draw. Do you remember the classic tutorial where step 1, you start with a bunch of circles, and step 2, you draw the rest of the fine owl? Now, you can actually have all the intermediate steps. Or, you know what? Just go the other way around. Get something that is done, and make a coloring book out of it. Four, it can also create maps. For instance, a map of Paper Land. As a young researcher, I spent years and years in dissertation desert… I survived by eating dry papers and filtered my tears into coffee. And when I go to the GTC AI conference and they ask what I did there, take a guess where I spent most of the time. Ha! I cannot stop making these silly little maps. You should do it too! So good. Five, if you see a piece of merch somewhere and you are thinking what you would look like wearing it, just ask. It can do it in a matter of seconds. Six, remember Gaugan? This was the state of the art just a few years ago. It was so limited in resolution, not a lot of detail there and it only understood a few semantic categories, like water, grass, sky and so on. So what can we do today in this direction? Is it any better? Dear Fellow Scholars, this is Two Minute Papers with Dr. Károly Zsolnai-Fehér. Well today, look at this. Wow. Making an image out of this would require an orders of magnitude better understanding of the world. And then, with the new ChatGPT, this happens. My goodness! This looks like something straight out of a science fiction movie. In fact, probably even better because most of them just keep doing the enhance thing. This is way better. Now hold on to your papers Fellow Scholars, for my favorite, number 7. Draw yourself a texture, and then ask it to create a teapot with this style. Now wait a minute. Hold on for a second. This is something that we call neural rendering, and when we wrote one of the earlier papers on it, we found that it works well, the only problem was that it took me 3,000 hours to research and implement it, and it was still limited. And now, you are telling me that this thing can just do it like that? With one prompt? Yes, that is exactly what I am saying. And it can even give you a different viewpoint, normal maps, or a full scene with beautiful glossy reflections and all. Is it perfect? Not even close. But just think about what we will be capable of just two more papers down the line. Now number 8. Yup, that is what I want. Please assemble this Ikea furniture for me and show me what my living room would look like if I ever got the patience to put it together. See? Not a problem. Now if only I could ask a robot
Segment 2 (05:00 - 05:00)
to actually go and do it by itself, that would be truly something else. So, 8 fantastic ideas for you, which one is your favorite? How did your Scholarly experiments go? Let me know in the comments below.