AI

AI has become a key factor in driving innovation across many sectors, and its impact on personal computing is now undeniable. Since the launch of its first RTX GPU in 2018, NVIDIA has been at the forefront of integrating AI into various applications. These GPUs feature Tensor Cores, dedicated AI hardware, and introduced technologies like Deep Learning Super Sampling (DLSS), which uses AI to enhance gaming visuals by generating pixels and frames.

Today, AI is embedded in countless aspects of PC use, including content creation, video streaming, video conferencing, and productivity tools. More than 600 million AI-enabled RTX-powered PCs are now installed globally, with over 600 AI applications accelerated by RTX GPUs. However, as AI technology continues to advance, its capability and accessibility have reached new heights, opening up exciting opportunities for developers and users alike.


A major shift in how software is written is underway. Traditional methods of software development involve developers writing code, which is then compiled into instructions executed primarily by the CPU. While effective for many tasks, this approach struggles to scale with today's complex challenges.

AI, particularly through machine learning, uses vast amounts of data to train neural networks that run on GPUs. This method is more flexible, adaptable, and better suited for dealing with complexity and new scenarios. Instead of relying on static instructions, AI-based software learns from billions of data samples to solve problems in a more dynamic and scalable way.

The introduction of generative AI has made software development even more accessible. With low-code and no-code tools, developers can now connect various AI functions through easy-to-use APIs that support text, images, 3D, and speech. These innovations allow a much broader audience—ranging from seasoned developers to creators, makers, and students—to dive into AI development.


To address the challenges of AI development, NVIDIA introduced NIM on RTX, a solution that simplifies the process for developers to leverage AI on the PC. NIM (NVIDIA Inference Models) microservices are prepackaged, optimized AI models designed to run on RTX GPUs. From what I understand, these are Docker containers (for people familiar with the Linux concept). These models are readily available for download, are easy to integrate with a variety of applications, and offer a range of features designed to make AI development on the PC as seamless as possible. Each NIM microservice includes optimized models, including both popular community-driven models and those developed by NVIDIA. These models are fine-tuned for RTX GPUs and are packaged in pre-built containers that eliminate the need for developers to handle the complex steps of model optimization, adaptation, and integration.


NVIDIA plans to launch the first wave of NIM microservices in February, spanning various modalities that are useful for PC development. These models will be available for free, with permissive licenses that allow developers to use, modify, and deploy them on RTX-powered PCs. This initiative marks a major step in democratizing AI development, giving anyone—from hobbyists to professionals—the tools they need to start building with AI.


NVIDIA also supports a wide range of AI tools, including no-code and low-code solutions, such as Crew.AI, ComfyUI, and Flow Wise.AI. These tools simplify the process of building AI applications and are compatible with NIM microservices, ensuring that developers can quickly experiment and iterate on their ideas.


At the core of what NVIDIA envisions as "the RTX AI PC" is the seamless integration of AI workloads into Windows, made possible by WSL (Windows Subsystem for Linux). Traditionally, AI development has been predominantly done in Linux environments, with microservices like NVIDIA's NIM running on Linux. The use of WSL now enables these workloads to be handled efficiently within Windows, allowing developers to run the same applications both locally and in the cloud. This offers significant flexibility for deploying AI workloads in hybrid scenarios, where some tasks run locally while others are offloaded to the cloud.

In line with these technological enhancements, NVIDIA introduced AI blueprints for RTX. These customizable reference projects serve as a starting point for developers, providing them with essential tools, source code, sample data, and a complete sample application to create AI-driven experiences. The blueprints allow developers to quickly get started, extending or building upon the base projects to suit their needs. Several of these blueprints were showcased, highlighting the potential of the new platform for AI innovation.


The "PDF to Podcast" demo demonstrated how the extracted content, once converted into audio, can be customized by selecting different podcast themes, such as science, technology, or health. This level of personalization allows users to tailor content to their specific interests or audiences. The ease of use was evident, with the system providing a smooth experience where a user can simply drag and drop a PDF file and let the AI do the rest.


In a live demonstration, a digital human avatar, known as R2X, showcased its abilities in a variety of real-world scenarios. R2X acted as a virtual assistant, helping users with tasks such as summarizing documents, assisting in meetings via Microsoft Teams, and even providing detailed instructions for editing images in Photoshop.

For instance, when asked how to replace a jacket in a photo, R2X guided the user through Photoshop's generative fill feature, illustrating how AI can streamline creative processes. In another scenario, the avatar helped analyze an insurance policy, answering questions about coverage in response to a potential snow-related roof leak.

For the future, NVIDIA mentioned that they are working with Adobe to give R2X direct access to Photoshop features, so that it can activate them directly, without requesting the interaction from you.

The demonstration also highlighted R2X's ability to provide real-time information, such as the date of NVIDIA's ray tracing panel and even the location of the nearest Starbucks.


For developers and AI enthusiasts eager to explore these new capabilities, NVIDIA has made it simple to get started. The AI blueprints and NIM (NVIDIA Integrated Models) will be accessible via NVIDIA's web-based playground, where users can test out different configurations and functionalities. Additionally, a one-click installer will soon be available, streamlining the setup process for those using RTX-powered PCs.