In this episode of "The Good Fight," Yascha Mounk interviews David Bau, a computer scientist from Northeastern University and former Google employee, to provide a 101 introduction to how AI models work. Bau explains the differences between traditional AI classifiers and modern large language models (LLMs), detailing the technology behind neural networks, neurons, and the training processes involved. He discusses the significance of the transformer architecture and its role in enabling short-term memory and contextual understanding in AI. Bau also addresses the two-step process of modern machine learning: pre-training and fine-tuning, and the distinction between supervised and unsupervised training methods. He also voices his concern about the trend of accepting AI as a black box without fully understanding its inner workings, advocating for more research into the interpretability of these complex systems.
Sign in to continue reading, translating and more.
Continue