This week on AI Decoded, the spotlight shines on the latest AI developments, broadcasting directly from the bustling floors of GTC, NVIDIA’s prestigious annual global AI conference.
ChatRTX (previously known as Chat with RTX) captured the attention of GTC attendees with its demonstration of enhanced capabilities and LLM (Large Language Model) integrations, promising future updates that will enrich the experience for over 100 million RTX AI PCs and workstations users worldwide. For those needing a refresher on LLMs, refer back to an earlier blog post.
The enhanced ChatRTX is set to revolutionize interaction by understanding spoken language and replying in text form, thanks to the Whisper AI automatic speech recognition system. It will also enable users to effortlessly locate photos in their personal collections using verbal requests, making the search for a two-decade-old picture as simple as pie, without the hassle of navigating through countless folders. Additionally, the inclusion of more LLM models like ChatGLM2 and Google’s Gemma is anticipated, offering greater versatility to users.
Other highlights from GTC include the further integration of RTX AI acceleration into creative applications such as those by Blackmagic Design and Topaz Labs, along with the expansion of TensorRT-LLM integration across popular developer frameworks like Jan.ai, Langchain, LlamaI-index, and Oobabooga.
NVIDIA is also set to introduce NIM inference microservices to RTX PCs, aiming to assist developers in creating an array of new AI applications, adding to the already impressive roster of over 500 RTX AI apps. For an in-depth understanding of these developments, AI Decoded breaks down the intricacies in this week’s episode.
The episode also highlights the three victors of the NVIDIA Generative AI on NVIDIA RTX developer contest. The winning entries range from an app enhancing Outlook with private generative AI capabilities to a project aimed at upgrading the PowerShell command line interface by translating simple English instructions into executable commands.
For those keen on delving deeper into AI, AI Decoded recommends watching the GTC keynote by NVIDIA CEO, Jensen Huang.