NVIDIA's Generative AI: Revolutionizing Robotics

0:00Generative AI has been kind of a focal point in the robotics world, which is there's a ton of really intense discussions about its applications and those are ranging from natural language language. inference to design. And then I guess essentially during a recent visit to NVIDIA's South Bay headquarters, there's a handful of journalists that spoke to Deepu Tala, who's the vice president and general manager of embedded and edge computing. And Tala was talking a lot about kind of this whole conversation and this whole topic of, you know, AI and robotics and what NVIDIA is doing at the moment. So he recently said, quote, I think it speaks in the results.

0:36You can already see the productivity improvements. It can compose an email for me. It's not exactly right, but I don't have to start from zero. It's giving me 70%. There are obvious things you can already see that are definitely a step function better than how things were before. So summarizing something's not perfect. I'm not going to let it read and summarize for me. So you can already see some signs of productivity improvements. I think notably NVIDIA was kind of on the verge of unveiling their groundbreaking developments related to generative AI. The ROS con announcement coincided with several other significant updates in the robotics offering.

1:12So I think this includes the general availability of NVIDIA ISAQ ROS 2.0 and NVIDIA ISAQ SIM 2023 platforms. So these systems are now embracing generative AI, a move that I think is expected to kind of like speed up its whole adoption around in robotics. NVIDIA points out that a staggering 1.2 million developers have engaged with NVIDIA AI and Jetson platforms. And they're counting, you know, heavyweight clients like AWS, Cisco, and John Deere that are actively working with them. So a highlight, I think, of this development is that Jetson Generative AI Lab, which is a hub that provides developers with access to open source large language models, in relation to all of this, NVIDIA recently said, quote, the NVIDIA Jetson Generative AI Lab provides developers access to optimized tools and tutorials for deploying open source LLMs, diffusion models to generate stunning images interactively, vision language models, and vision transformers that combine vision AI and natural language processors to provide comprehensive understanding of the scene.

2:21So I think these models are really kind of poised to enhance the capabilities of systems to make decisions in previously untrained scenarios, particularly critical in dynamic environments like warehouses and factory floors. So the idea is essentially to enable these systems to adapt on the fly while offering kind of a more, you know, a more natural language interface with all of them. So Deeputala was kind of underlying this in some recent comments saying the significance of generative AI in this context is that, quote, generative AI will significantly accelerate deployment of AI at the edge with better generalization, ease of use, and higher accuracy than previously possible.

3:02The largest ever software expansion of our Metropolis and AI frameworks on Jetson combined with the power of transformer models in generative AI addresses this need. I think in addition to all of these advancements that we're seeing in AI and robotics today, you know, the latest version of these platforms also brings improvements in perception simulation, which is really kind of bringing in a whole new era where generative AI plays a really pivotal role in shaping the future of robotics. We've definitely already seen a lot of big robotics revolutions. You could just look at the fact that Amazon, for example, and all the Amazon warehouses, some of these things are like insanely automated with robotics.

3:39There's not even a lot of people on the floor where essentially there's just like shelves and shelves all over the place full of different products. And there's all these robots that just zip around, grab whatever you need, bring it to the fulfillment area just in time. That's how they've kind of got this whole like one day or two day shipping thing going really well in the US is because of robotics and how they're able to automate and kind of run their warehouses. Now, what I think is going to be really interesting is when these robotics really get a lot more AI embedded into them. And of course, there's something like an Amazon warehouse, but I think there's a lot, there's a lot of really interesting use cases in my mind, beyond just maybe robots that are doing very like repetitive tasks.

4:18So you can imagine, for example, like a robot that is an electrician that's trained to, you know, it's like a humanoid robot that can climb up an electrical pole and get something hooked up that, you know, was previously sort of dangerous for a human to do. So I think there's a lot of really exciting use cases here. I think NVIDIA is definitely, they're building software, they're building tools, it looks like they're building chips and other things that are going to be really useful in this space and so i think nvidia is definitely a company that's not missing out on this and probably one that's going to continue to grow when you kind of look at companies that are very forward thinking to where this whole industry goes nvidia would appear to be one of those companies it's going to be really interesting to see how some of the software tools they're building enable ai to go into these robotics definitely a story will continue to follow in the future