Finding Nemotron
Practical AI
2025/07/02
Finding Nemotron
Finding Nemotron
Practical AI
2025/07/02
In this episode, the hosts of Practical AI sit down with Joey Conway from NVIDIA to discuss the evolving landscape of open-source AI models and their impact on enterprise applications. The conversation delves into NVIDIA's strategic balance between open and closed AI systems, the company's long-standing commitment to advancing AI innovation, and the growing importance of reasoning capabilities in generative AI.
Joey Conway highlights NVIDIA's development of reasoning-focused models like the Nemotron series, built on top of Llama, and the high-performance Parakeet speech model. The discussion covers how enterprises can strategically adopt multi-model approaches, evaluating models like Llama 3.1, Gemini, or ChatGPT based on their unique strengths and suitability for specific tasks. Conway emphasizes the importance of reasoning in AI, explaining how models are now capable of breaking down complex problems and offering multiple solutions. The episode also explores technical advancements, including parameter scaling strategies, architectural optimizations, and performance enhancements in speech recognition. NVIDIA's open-source initiatives and partnerships are noted as key drivers for broader AI adoption and innovation across languages and industries.
03:49
03:49
NVIDIA publishes models and datasets to support AI infrastructure and community innovation
12:10
12:10
Training models to break down complex problems with reasoning traces
21:13
21:13
Published a family of three models (nano, super, ultra) with different scales and capabilities
26:33
26:33
Close collaboration with Meta confirmed, joint achievements celebrated at LlamaCon
34:21
34:21
Nemotron excels in enterprise agentic tasks with complex tool calling and multi-source reasoning.
37:04
37:04
Parakeet speech model achieves 2-3x inference speed-up with architectural improvements
44:12
44:12
Digital agents will transform business by handling repetitive tasks
