GTC 2021 Keynote

 GTC 2021 conference is happening this week virtually.  Looking forwards to Jurgen's talk.  The other 3 'fathers of deep learning (ie: Turing Award winners)' are also speaking.

Nvidia CEO Jenson Huang once again addresses us from his kitchen, schooling us on the latest developments in Nvidia GPUs, graphics and AI technology.  Feel free to skip the 18 minute commercial at the beginning before he comes on live.



Observations

1:  Are we really living in the 'metaverse'?  Neil Stephensons's Snow Crash novel is a story about a dystopia at it's core.  Surely we can come up with something better as a metaphor to strive for.

2: Omniverse is cool as a tool kit. Keep meaning to look into using the api for something more artistic.

3: Interesting how deep learning is now the key driver of increased GPU services in the cloud.

4:  Look at all those ARM cores on that Bluefield 3-3 DPU. So this is where we continue Moore's law growth? 

Now please stick DGX superpod in a small box for my personal computer (ha, ha, but seriously, please).

5:  Be very careful comparing Transformer architectures to how the brain works.

6:  The computational chemistry, genetics, and bio-design stuff is really great.

7:  Yes, that small x86 box in the memory - x86- GPU slide is indeed the bottleneck. So throw it out in the trash and put in that wide stack of ARM Grace blocks.

Now please stick this architecture in a small box to be my personal computer (ha, ha, but seriously, please).

So is the MediaTech initiative going to give me what i'm looking for?

8: Software 2.0 metaphor reference (several actually).

Nvidia AI works on any system (except apple, ha,ha, groan, frustrated sigh).

9:  NGC Pre-Trained deep learning models.

Tao software for transfer learning and deployment quantization of the customized model.

10: The AI voice we use for our new Studio Artist training videos sounds way better than Jarvis sounds in this keynote.

11: How does apple's secret over-hyped car stuff compete with Atlan (with it's associated vast AI infrastructure)?


Jenson always looks like he's ready to tour with Front 242 with the black leather jacket (which i actually can appreciate, since it's the old OSC aesthetic).

Comments

Post a Comment

Popular posts from this blog

CycleGAN: a GAN architecture for learning unpaired image to image transformations

Image
 Like the Doublemint Twins touting the joys of Doublemint gum, 2 GANs are surely better than 1 GAN.  Especially if we package them together inside of one meta GAN module.  And this is exactly what the CycleGAN architecture does. Have you ever harbored dark secrets of turning a horse into a zebra?  The CycleGAN was developed to do just that. Learn how to turn a horse into a zebra.  And more. Now right away you can notice a difference between the image to image transformation GAN architectures we've been discussing over the last few posts.  Those last few posts described systems that learn from a database of matched input-output image pairs. And if your goal is to turn an edge representation into a nicely filled in continuous tone image, it's easy to build your database of matched input-output image pairs that your GAN system can then learn off of.  Take a continuous tone image (which will be the output of the database pair entry), then run it through an edge detector algorithm. 
Read more

Pix2Pix: a GAN architecture for image to image transformation

Image
 I thought following up yesterday's TraVelGAN post with a Pix2Pix GAN post would be useful to compare what is going on in the 2 architectures.  Two different approaches to the same problem. I stole this Pix2Pix Overview slide below from an excellent deeplearning.ai GAN course (note that they borrowed it from the original paper) because it gives you a good feel for what is going on inside of the Pix2Pix architecture.    Note how the Generator part is very much like an auto-encoder architecture, but rebuilt using the U-Net architecture features (based on skip-connections) that fastai has been discussing in their courses for several years before it became more widely known to the deep learning community at large  (and which originally came from an obscure medical image segmentation paper) . So the Generator in this Pix2Pix GAN is really pretty sophisticated, consisting of a whole image to image auto-encoder network with U-Net skip connections to generate better image quality at highe
Read more

Smart Fabrics

Image
"What makes smart fabrics revolutionary is that they have the ability to do many things that traditional fabrics cannot, including communicate, transform, conduct energy and even grow."  This is a quote from Rebeccah Pailes-Friedman, who is a professor at Pratt with a research focus on developing wearable technology and smart textiles. Enhancing textiles can be viewed through different lens.  Enhancements can be aesthetic, or performance boosting, or durability boosting. 
Read more