Computational Bottlenecks of Training Small-scale Large Language Models | AI Papers Podcast Daily | Podwise