This transcript discusses how programming languages can be a barrier to implementing cutting-edge research in deep learning, and suggests that higher-level languages like Swift could allow for easier experimentation with neural network architectures like RNNs and convolutional neural networks.