In this episode of the ‘Deep Learning with PolyAI’ podcast, host Damien speaks with Tomasz, a senior machine learning engineer at PolyAI, about his experiences at NVIDIA’s GTC 2024. They discuss a few standout sessions, including the large swathe on Large Language Models (LLMs), generative AI, and the NVIDIA Blackwater platform, which aims to significantly reduce the energy and cost requirements of high-performance AI systems.

Tomasz also shares insights on the discussions around scaling infrastructure, responsible AI utilization, and the shift towards smaller, quality-focused data models over larger ones. Additionally, they touch on Poly AI’s presentation on Automatic Speech Recognition (ASR) technology and its implications for spoken language understanding. The episode concludes with Tomasz providing observations on emerging trends and potential industry shifts highlighted during the conference.