In The Neuron podcast, Kwasi Ankomah, the lead AI architect at SambaNova Systems, discusses AI inference and agentic AI. He explains inference as the process of a model predicting the next token and highlights the challenges of making inference fast and scalable for real-world applications. Kwasi details SambaNova's approach, which involves custom chips and platforms that efficiently run large AI models, emphasizing the importance of speed, cost, and energy consumption. He also touches on the developer experience with SambaNova, the agentic use of their platform, and the future of AI with a mix of both large and niche models. The discussion also covers the significance of power efficiency and the potential of open-source models in AI development.
Sign in to continue reading, translating and more.
Continue